open-consul/agent/consul/acl_endpoint_test.go

5739 lines
152 KiB
Go
Raw Normal View History

2014-08-06 00:24:48 +00:00
package consul
import (
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
"fmt"
"io/ioutil"
"net/rpc"
2014-08-06 00:24:48 +00:00
"os"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
"path/filepath"
"reflect"
2014-08-12 22:32:44 +00:00
"strings"
2014-08-06 00:24:48 +00:00
"testing"
2014-08-08 22:32:43 +00:00
"time"
"github.com/hashicorp/consul/acl"
"github.com/hashicorp/consul/agent/consul/authmethod/kubeauth"
"github.com/hashicorp/consul/agent/consul/authmethod/testauth"
"github.com/hashicorp/consul/agent/structs"
ACL Token Persistence and Reloading (#5328) This PR adds two features which will be useful for operators when ACLs are in use. 1. Tokens set in configuration files are now reloadable. 2. If `acl.enable_token_persistence` is set to `true` in the configuration, tokens set via the `v1/agent/token` endpoint are now persisted to disk and loaded when the agent starts (or during configuration reload) Note that token persistence is opt-in so our users who do not want tokens on the local disk will see no change. Some other secondary changes: * Refactored a bunch of places where the replication token is retrieved from the token store. This token isn't just for replicating ACLs and now it is named accordingly. * Allowed better paths in the `v1/agent/token/` API. Instead of paths like: `v1/agent/token/acl_replication_token` the path can now be just `v1/agent/token/replication`. The old paths remain to be valid. * Added a couple new API functions to set tokens via the new paths. Deprecated the old ones and pointed to the new names. The names are also generally better and don't imply that what you are setting is for ACLs but rather are setting ACL tokens. There is a minor semantic difference there especially for the replication token as again, its no longer used only for ACL token/policy replication. The new functions will detect 404s and fallback to using the older token paths when talking to pre-1.4.3 agents. * Docs updated to reflect the API additions and to show using the new endpoints. * Updated the ACL CLI set-agent-tokens command to use the non-deprecated APIs.
2019-02-27 19:28:31 +00:00
tokenStore "github.com/hashicorp/consul/agent/token"
"github.com/hashicorp/consul/lib"
"github.com/hashicorp/consul/sdk/testutil/retry"
"github.com/hashicorp/consul/testrpc"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
uuid "github.com/hashicorp/go-uuid"
msgpackrpc "github.com/hashicorp/net-rpc-msgpackrpc"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
"github.com/stretchr/testify/require"
2014-08-06 00:24:48 +00:00
)
func TestACLEndpoint_Bootstrap(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.Build = "0.8.0" // Too low for auto init of bootstrap.
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
// Expect an error initially since ACL bootstrap is not initialized.
arg := structs.DCSpecificRequest{
Datacenter: "dc1",
}
var out structs.ACL
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// We can only do some high
// level checks on the ACL since we don't have control over the UUID or
// Raft indexes at this level.
if err := msgpackrpc.CallWithCodec(codec, "ACL.Bootstrap", &arg, &out); err != nil {
t.Fatalf("err: %v", err)
}
if len(out.ID) != len("xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx") ||
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
!strings.HasPrefix(out.Name, "Bootstrap Token") ||
out.Type != structs.ACLTokenTypeManagement ||
out.CreateIndex == 0 || out.ModifyIndex == 0 {
t.Fatalf("bad: %#v", out)
}
// Finally, make sure that another attempt is rejected.
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
err := msgpackrpc.CallWithCodec(codec, "ACL.Bootstrap", &arg, &out)
if err.Error() != structs.ACLBootstrapNotAllowedErr.Error() {
t.Fatalf("err: %v", err)
}
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
func TestACLEndpoint_BootstrapTokens(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLsEnabled = true
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
// Expect an error initially since ACL bootstrap is not initialized.
arg := structs.DCSpecificRequest{
Datacenter: "dc1",
}
var out structs.ACLToken
// We can only do some high
// level checks on the ACL since we don't have control over the UUID or
// Raft indexes at this level.
require.NoError(t, msgpackrpc.CallWithCodec(codec, "ACL.BootstrapTokens", &arg, &out))
require.Equal(t, 36, len(out.AccessorID))
require.True(t, strings.HasPrefix(out.Description, "Bootstrap Token"))
require.Equal(t, out.Type, structs.ACLTokenTypeManagement)
require.True(t, out.CreateIndex > 0)
require.Equal(t, out.CreateIndex, out.ModifyIndex)
// Finally, make sure that another attempt is rejected.
err := msgpackrpc.CallWithCodec(codec, "ACL.BootstrapTokens", &arg, &out)
require.Error(t, err)
require.True(t, strings.HasPrefix(err.Error(), structs.ACLBootstrapNotAllowedErr.Error()))
_, resetIdx, err := s1.fsm.State().CanBootstrapACLToken()
resetPath := filepath.Join(dir1, "acl-bootstrap-reset")
require.NoError(t, ioutil.WriteFile(resetPath, []byte(fmt.Sprintf("%d", resetIdx)), 0600))
oldID := out.AccessorID
// Finally, make sure that another attempt is rejected.
require.NoError(t, msgpackrpc.CallWithCodec(codec, "ACL.BootstrapTokens", &arg, &out))
require.Equal(t, 36, len(out.AccessorID))
require.NotEqual(t, oldID, out.AccessorID)
require.True(t, strings.HasPrefix(out.Description, "Bootstrap Token"))
require.Equal(t, out.Type, structs.ACLTokenTypeManagement)
require.True(t, out.CreateIndex > 0)
require.Equal(t, out.CreateIndex, out.ModifyIndex)
}
2014-08-06 00:24:48 +00:00
func TestACLEndpoint_Apply(t *testing.T) {
t.Parallel()
2014-08-12 22:32:44 +00:00
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
2014-08-12 22:32:44 +00:00
c.ACLMasterToken = "root"
})
2014-08-06 00:24:48 +00:00
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
2014-08-06 00:24:48 +00:00
testrpc.WaitForLeader(t, s1.RPC, "dc1")
2014-08-06 00:24:48 +00:00
arg := structs.ACLRequest{
Datacenter: "dc1",
Op: structs.ACLSet,
ACL: structs.ACL{
Name: "User token",
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Type: structs.ACLTokenTypeClient,
2014-08-06 00:24:48 +00:00
},
2014-08-12 22:32:44 +00:00
WriteRequest: structs.WriteRequest{Token: "root"},
2014-08-06 00:24:48 +00:00
}
var out string
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out); err != nil {
2014-08-06 00:24:48 +00:00
t.Fatalf("err: %v", err)
}
id := out
// Verify
state := s1.fsm.State()
_, s, err := state.ACLTokenGetBySecret(nil, out, nil)
2014-08-06 00:24:48 +00:00
if err != nil {
t.Fatalf("err: %v", err)
}
if s == nil {
t.Fatalf("should not be nil")
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
if s.SecretID != out {
2014-08-06 00:24:48 +00:00
t.Fatalf("bad: %v", s)
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
if s.Description != "User token" {
2014-08-06 00:24:48 +00:00
t.Fatalf("bad: %v", s)
}
// Do a delete
arg.Op = structs.ACLDelete
arg.ACL.ID = out
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out); err != nil {
2014-08-06 00:24:48 +00:00
t.Fatalf("err: %v", err)
}
// Verify
_, s, err = state.ACLTokenGetBySecret(nil, id, nil)
2014-08-06 00:24:48 +00:00
if err != nil {
t.Fatalf("err: %v", err)
}
if s != nil {
t.Fatalf("bad: %v", s)
}
}
func TestACLEndpoint_Update_PurgeCache(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
arg := structs.ACLRequest{
Datacenter: "dc1",
Op: structs.ACLSet,
ACL: structs.ACL{
Name: "User token",
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Type: structs.ACLTokenTypeClient,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
var out string
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out); err != nil {
t.Fatalf("err: %v", err)
}
id := out
// Resolve
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
acl1, err := s1.ResolveToken(id)
if err != nil {
t.Fatalf("err: %v", err)
}
if acl1 == nil {
t.Fatalf("should not be nil")
}
if acl1.KeyRead("foo", nil) != acl.Allow {
t.Fatalf("should be allowed")
}
// Do an update
arg.ACL.ID = out
arg.ACL.Rules = `{"key": {"": {"policy": "deny"}}}`
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out); err != nil {
t.Fatalf("err: %v", err)
}
// Resolve again
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
acl2, err := s1.ResolveToken(id)
if err != nil {
t.Fatalf("err: %v", err)
}
if acl2 == nil {
t.Fatalf("should not be nil")
}
if acl2 == acl1 {
t.Fatalf("should not be cached")
}
if acl2.KeyRead("foo", nil) == acl.Allow {
t.Fatalf("should not be allowed")
}
// Do a delete
arg.Op = structs.ACLDelete
arg.ACL.Rules = ""
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out); err != nil {
t.Fatalf("err: %v", err)
}
// Resolve again
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
acl3, err := s1.ResolveToken(id)
if !acl.IsErrNotFound(err) {
t.Fatalf("err: %v", err)
}
if acl3 != nil {
t.Fatalf("should be nil")
}
}
func TestACLEndpoint_Apply_CustomID(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
arg := structs.ACLRequest{
Datacenter: "dc1",
Op: structs.ACLSet,
ACL: structs.ACL{
ID: "foobarbaz", // Specify custom ID, does not exist
Name: "User token",
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Type: structs.ACLTokenTypeClient,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
var out string
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out); err != nil {
t.Fatalf("err: %v", err)
}
if out != "foobarbaz" {
t.Fatalf("bad token ID: %s", out)
}
// Verify
state := s1.fsm.State()
_, s, err := state.ACLTokenGetBySecret(nil, out, nil)
if err != nil {
t.Fatalf("err: %v", err)
}
if s == nil {
t.Fatalf("should not be nil")
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
if s.SecretID != out {
t.Fatalf("bad: %v", s)
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
if s.Description != "User token" {
t.Fatalf("bad: %v", s)
}
}
2014-08-12 22:32:44 +00:00
func TestACLEndpoint_Apply_Denied(t *testing.T) {
t.Parallel()
2014-08-12 22:32:44 +00:00
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
2014-08-12 22:32:44 +00:00
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
2014-08-12 22:32:44 +00:00
testrpc.WaitForLeader(t, s1.RPC, "dc1")
2014-08-12 22:32:44 +00:00
arg := structs.ACLRequest{
Datacenter: "dc1",
Op: structs.ACLSet,
ACL: structs.ACL{
Name: "User token",
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Type: structs.ACLTokenTypeClient,
2014-08-12 22:32:44 +00:00
},
}
var out string
err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out)
if !acl.IsErrPermissionDenied(err) {
2014-08-12 22:32:44 +00:00
t.Fatalf("err: %v", err)
}
}
func TestACLEndpoint_Apply_DeleteAnon(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
arg := structs.ACLRequest{
Datacenter: "dc1",
Op: structs.ACLDelete,
ACL: structs.ACL{
ID: anonymousToken,
Name: "User token",
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Type: structs.ACLTokenTypeClient,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
var out string
err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out)
if err == nil || !strings.Contains(err.Error(), "delete anonymous") {
t.Fatalf("err: %v", err)
}
}
func TestACLEndpoint_Apply_RootChange(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
arg := structs.ACLRequest{
Datacenter: "dc1",
Op: structs.ACLSet,
ACL: structs.ACL{
ID: "manage",
Name: "User token",
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Type: structs.ACLTokenTypeClient,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
var out string
err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out)
if err == nil || !strings.Contains(err.Error(), "root ACL") {
t.Fatalf("err: %v", err)
}
}
2014-08-06 00:24:48 +00:00
func TestACLEndpoint_Get(t *testing.T) {
t.Parallel()
2014-08-12 22:32:44 +00:00
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
2014-08-12 22:32:44 +00:00
c.ACLMasterToken = "root"
})
2014-08-06 00:24:48 +00:00
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
2014-08-06 00:24:48 +00:00
testrpc.WaitForLeader(t, s1.RPC, "dc1")
2014-08-06 00:24:48 +00:00
arg := structs.ACLRequest{
Datacenter: "dc1",
Op: structs.ACLSet,
ACL: structs.ACL{
Name: "User token",
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Type: structs.ACLTokenTypeClient,
2014-08-06 00:24:48 +00:00
},
2014-08-12 22:32:44 +00:00
WriteRequest: structs.WriteRequest{Token: "root"},
2014-08-06 00:24:48 +00:00
}
var out string
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out); err != nil {
2014-08-06 00:24:48 +00:00
t.Fatalf("err: %v", err)
}
getR := structs.ACLSpecificRequest{
Datacenter: "dc1",
ACL: out,
}
var acls structs.IndexedACLs
if err := msgpackrpc.CallWithCodec(codec, "ACL.Get", &getR, &acls); err != nil {
2014-08-06 00:24:48 +00:00
t.Fatalf("err: %v", err)
}
if acls.Index == 0 {
t.Fatalf("Bad: %v", acls)
}
if len(acls.ACLs) != 1 {
t.Fatalf("Bad: %v", acls)
}
s := acls.ACLs[0]
if s.ID != out {
t.Fatalf("bad: %v", s)
}
}
2014-08-08 22:32:43 +00:00
func TestACLEndpoint_GetPolicy(t *testing.T) {
t.Parallel()
2014-08-12 22:32:44 +00:00
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
2014-08-12 22:32:44 +00:00
c.ACLMasterToken = "root"
})
2014-08-08 22:32:43 +00:00
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
2014-08-08 22:32:43 +00:00
testrpc.WaitForLeader(t, s1.RPC, "dc1")
2014-08-08 22:32:43 +00:00
arg := structs.ACLRequest{
Datacenter: "dc1",
Op: structs.ACLSet,
ACL: structs.ACL{
Name: "User token",
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Type: structs.ACLTokenTypeClient,
2014-08-08 22:32:43 +00:00
},
2014-08-12 22:32:44 +00:00
WriteRequest: structs.WriteRequest{Token: "root"},
2014-08-08 22:32:43 +00:00
}
var out string
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out); err != nil {
2014-08-08 22:32:43 +00:00
t.Fatalf("err: %v", err)
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
getR := structs.ACLPolicyResolveLegacyRequest{
2014-08-08 22:32:43 +00:00
Datacenter: "dc1",
ACL: out,
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
var acls structs.ACLPolicyResolveLegacyResponse
retry.Run(t, func(r *retry.R) {
if err := msgpackrpc.CallWithCodec(codec, "ACL.GetPolicy", &getR, &acls); err != nil {
r.Fatalf("err: %v", err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
if acls.Policy == nil {
r.Fatalf("Bad: %v", acls)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
if acls.TTL != 30*time.Second {
r.Fatalf("bad: %v", acls)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
})
// Do a conditional lookup with etag
getR.ETag = acls.ETag
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
var out2 structs.ACLPolicyResolveLegacyResponse
if err := msgpackrpc.CallWithCodec(codec, "ACL.GetPolicy", &getR, &out2); err != nil {
t.Fatalf("err: %v", err)
}
if out2.Policy != nil {
t.Fatalf("Bad: %v", out2)
}
if out2.TTL != 30*time.Second {
t.Fatalf("bad: %v", out2)
}
2014-08-08 22:32:43 +00:00
}
2014-08-06 00:24:48 +00:00
func TestACLEndpoint_List(t *testing.T) {
t.Parallel()
2014-08-12 22:32:44 +00:00
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
2014-08-12 22:32:44 +00:00
c.ACLMasterToken = "root"
})
2014-08-06 00:24:48 +00:00
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
2014-08-06 00:24:48 +00:00
testrpc.WaitForLeader(t, s1.RPC, "dc1")
2014-08-06 00:24:48 +00:00
ids := []string{}
for i := 0; i < 5; i++ {
arg := structs.ACLRequest{
Datacenter: "dc1",
Op: structs.ACLSet,
ACL: structs.ACL{
Name: "User token",
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Type: structs.ACLTokenTypeClient,
2014-08-06 00:24:48 +00:00
},
2014-08-12 22:32:44 +00:00
WriteRequest: structs.WriteRequest{Token: "root"},
2014-08-06 00:24:48 +00:00
}
var out string
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &out); err != nil {
2014-08-06 00:24:48 +00:00
t.Fatalf("err: %v", err)
}
ids = append(ids, out)
}
getR := structs.DCSpecificRequest{
2014-08-12 22:32:44 +00:00
Datacenter: "dc1",
QueryOptions: structs.QueryOptions{Token: "root"},
2014-08-06 00:24:48 +00:00
}
var acls structs.IndexedACLs
if err := msgpackrpc.CallWithCodec(codec, "ACL.List", &getR, &acls); err != nil {
2014-08-06 00:24:48 +00:00
t.Fatalf("err: %v", err)
}
if acls.Index == 0 {
t.Fatalf("Bad: %v", acls)
}
2014-08-12 22:32:44 +00:00
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// 5 + master
if len(acls.ACLs) != 6 {
2014-08-06 00:24:48 +00:00
t.Fatalf("Bad: %v", acls.ACLs)
}
for i := 0; i < len(acls.ACLs); i++ {
s := acls.ACLs[i]
2014-08-12 22:32:44 +00:00
if s.ID == anonymousToken || s.ID == "root" {
continue
}
if !lib.StrContains(ids, s.ID) {
2014-08-06 00:24:48 +00:00
t.Fatalf("bad: %v", s)
}
if s.Name != "User token" {
t.Fatalf("bad: %v", s)
}
}
}
2014-08-12 22:32:44 +00:00
func TestACLEndpoint_List_Denied(t *testing.T) {
t.Parallel()
2014-08-12 22:32:44 +00:00
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
2014-08-12 22:32:44 +00:00
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
2014-08-12 22:32:44 +00:00
testrpc.WaitForLeader(t, s1.RPC, "dc1")
2014-08-12 22:32:44 +00:00
getR := structs.DCSpecificRequest{
Datacenter: "dc1",
}
var acls structs.IndexedACLs
err := msgpackrpc.CallWithCodec(codec, "ACL.List", &getR, &acls)
if !acl.IsErrPermissionDenied(err) {
2014-08-12 22:32:44 +00:00
t.Fatalf("err: %v", err)
}
}
func TestACLEndpoint_ReplicationStatus(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc2"
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
c.ACLsEnabled = true
c.ACLTokenReplication = true
c.ACLReplicationRate = 100
c.ACLReplicationBurst = 100
})
ACL Token Persistence and Reloading (#5328) This PR adds two features which will be useful for operators when ACLs are in use. 1. Tokens set in configuration files are now reloadable. 2. If `acl.enable_token_persistence` is set to `true` in the configuration, tokens set via the `v1/agent/token` endpoint are now persisted to disk and loaded when the agent starts (or during configuration reload) Note that token persistence is opt-in so our users who do not want tokens on the local disk will see no change. Some other secondary changes: * Refactored a bunch of places where the replication token is retrieved from the token store. This token isn't just for replicating ACLs and now it is named accordingly. * Allowed better paths in the `v1/agent/token/` API. Instead of paths like: `v1/agent/token/acl_replication_token` the path can now be just `v1/agent/token/replication`. The old paths remain to be valid. * Added a couple new API functions to set tokens via the new paths. Deprecated the old ones and pointed to the new names. The names are also generally better and don't imply that what you are setting is for ACLs but rather are setting ACL tokens. There is a minor semantic difference there especially for the replication token as again, its no longer used only for ACL token/policy replication. The new functions will detect 404s and fallback to using the older token paths when talking to pre-1.4.3 agents. * Docs updated to reflect the API additions and to show using the new endpoints. * Updated the ACL CLI set-agent-tokens command to use the non-deprecated APIs.
2019-02-27 19:28:31 +00:00
s1.tokens.UpdateReplicationToken("secret", tokenStore.TokenSourceConfig)
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
getR := structs.DCSpecificRequest{
Datacenter: "dc1",
}
retry.Run(t, func(r *retry.R) {
var status structs.ACLReplicationStatus
err := msgpackrpc.CallWithCodec(codec, "ACL.ReplicationStatus", &getR, &status)
if err != nil {
r.Fatalf("err: %v", err)
}
if !status.Enabled || !status.Running || status.SourceDatacenter != "dc2" {
r.Fatalf("bad: %#v", status)
}
})
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
func TestACLEndpoint_TokenRead(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
t.Run("exists and matches what we created", func(t *testing.T) {
token, err := upsertTestToken(codec, "root", "dc1", nil)
require.NoError(t, err)
req := structs.ACLTokenGetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
TokenID: token.AccessorID,
TokenIDType: structs.ACLTokenAccessor,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLTokenResponse{}
err = acl.TokenRead(&req, &resp)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
if !reflect.DeepEqual(resp.Token, token) {
t.Fatalf("tokens are not equal: %v != %v", resp.Token, token)
}
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Run("expired tokens are filtered", func(t *testing.T) {
// insert a token that will expire
token, err := upsertTestToken(codec, "root", "dc1", func(t *structs.ACLToken) {
2019-07-12 15:52:26 +00:00
t.ExpirationTTL = 200 * time.Millisecond
})
require.NoError(t, err)
t.Run("readable until expiration", func(t *testing.T) {
req := structs.ACLTokenGetRequest{
Datacenter: "dc1",
TokenID: token.AccessorID,
TokenIDType: structs.ACLTokenAccessor,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLTokenResponse{}
require.NoError(t, acl.TokenRead(&req, &resp))
require.Equal(t, token, resp.Token)
})
t.Run("not returned when expired", func(t *testing.T) {
req := structs.ACLTokenGetRequest{
Datacenter: "dc1",
TokenID: token.AccessorID,
TokenIDType: structs.ACLTokenAccessor,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLTokenResponse{}
2019-07-12 15:52:26 +00:00
retry.Run(t, func(r *retry.R) {
require.NoError(r, acl.TokenRead(&req, &resp))
require.Nil(r, resp.Token)
})
})
})
t.Run("nil when token does not exist", func(t *testing.T) {
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
fakeID, err := uuid.GenerateUUID()
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
req := structs.ACLTokenGetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
TokenID: fakeID,
TokenIDType: structs.ACLTokenAccessor,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLTokenResponse{}
err = acl.TokenRead(&req, &resp)
require.Nil(t, resp.Token)
require.NoError(t, err)
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Run("validates ID format", func(t *testing.T) {
req := structs.ACLTokenGetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
TokenID: "definitely-really-certainly-not-a-uuid",
TokenIDType: structs.ACLTokenAccessor,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLTokenResponse{}
err := acl.TokenRead(&req, &resp)
require.Nil(t, resp.Token)
require.EqualError(t, err, "failed acl token lookup: index error: UUID must be 36 characters")
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_TokenClone(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
t1, err := upsertTestToken(codec, "root", "dc1", nil)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
endpoint := ACL{srv: s1}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Run("normal", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{AccessorID: t1.AccessorID},
WriteRequest: structs.WriteRequest{Token: "root"},
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t2 := structs.ACLToken{}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
err = endpoint.TokenClone(&req, &t2)
require.NoError(t, err)
require.Equal(t, t1.Description, t2.Description)
require.Equal(t, t1.Policies, t2.Policies)
require.Equal(t, t1.Rules, t2.Rules)
require.Equal(t, t1.Local, t2.Local)
require.NotEqual(t, t1.AccessorID, t2.AccessorID)
require.NotEqual(t, t1.SecretID, t2.SecretID)
})
t.Run("can't clone expired token", func(t *testing.T) {
// insert a token that will expire
t1, err := upsertTestToken(codec, "root", "dc1", func(t *structs.ACLToken) {
t.ExpirationTTL = 11 * time.Millisecond
})
require.NoError(t, err)
time.Sleep(30 * time.Millisecond)
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{AccessorID: t1.AccessorID},
WriteRequest: structs.WriteRequest{Token: "root"},
}
t2 := structs.ACLToken{}
err = endpoint.TokenClone(&req, &t2)
require.Error(t, err)
require.Equal(t, acl.ErrNotFound, err)
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_TokenSet(t *testing.T) {
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
var tokenID string
t.Run("Create it", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
token := tokenResp.Token
require.NotNil(t, token)
require.NotNil(t, token.AccessorID)
require.Equal(t, token.Description, "foobar")
require.Equal(t, token.AccessorID, resp.AccessorID)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
tokenID = token.AccessorID
})
t.Run("Update it", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "new-description",
AccessorID: tokenID,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
token := tokenResp.Token
require.NotNil(t, token)
require.NotNil(t, token.AccessorID)
require.Equal(t, token.Description, "new-description")
require.Equal(t, token.AccessorID, resp.AccessorID)
})
t.Run("Create it using Policies linked by id and name", func(t *testing.T) {
policy1, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
policy2, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: []structs.ACLTokenPolicyLink{
structs.ACLTokenPolicyLink{
ID: policy1.ID,
},
structs.ACLTokenPolicyLink{
Name: policy2.Name,
},
},
Local: false,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err = acl.TokenSet(&req, &resp)
require.NoError(t, err)
// Delete both policies to ensure that we skip resolving ID->Name
// in the returned data.
require.NoError(t, deleteTestPolicy(codec, "root", "dc1", policy1.ID))
require.NoError(t, deleteTestPolicy(codec, "root", "dc1", policy2.ID))
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
require.NotNil(t, token)
require.NotNil(t, token.AccessorID)
require.Equal(t, token.Description, "foobar")
require.Equal(t, token.AccessorID, resp.AccessorID)
require.Len(t, token.Policies, 0)
})
t.Run("Create it using Roles linked by id and name", func(t *testing.T) {
role1, err := upsertTestRole(codec, "root", "dc1")
require.NoError(t, err)
role2, err := upsertTestRole(codec, "root", "dc1")
require.NoError(t, err)
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Roles: []structs.ACLTokenRoleLink{
structs.ACLTokenRoleLink{
ID: role1.ID,
},
structs.ACLTokenRoleLink{
Name: role2.Name,
},
},
Local: false,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err = acl.TokenSet(&req, &resp)
require.NoError(t, err)
// Delete both roles to ensure that we skip resolving ID->Name
// in the returned data.
require.NoError(t, deleteTestRole(codec, "root", "dc1", role1.ID))
require.NoError(t, deleteTestRole(codec, "root", "dc1", role2.ID))
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
require.NotNil(t, token)
require.NotNil(t, token.AccessorID)
require.Equal(t, token.Description, "foobar")
require.Equal(t, token.AccessorID, resp.AccessorID)
require.Len(t, token.Roles, 0)
})
t.Run("Create it with AuthMethod set outside of login", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
AuthMethod: "fakemethod",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
requireErrorContains(t, err, "AuthMethod field is disallowed outside of Login")
})
t.Run("Update auth method linked token and try to change auth method", func(t *testing.T) {
acl := ACL{srv: s1}
testSessionID := testauth.StartSession()
defer testauth.ResetSession(testSessionID)
testauth.InstallSessionToken(testSessionID, "fake-token", "default", "demo", "abc123")
method1, err := upsertTestAuthMethod(codec, "root", "dc1", testSessionID)
require.NoError(t, err)
_, err = upsertTestBindingRule(codec, "root", "dc1", method1.Name, "", structs.BindingRuleBindTypeService, "demo")
require.NoError(t, err)
// create a token in one method
methodToken := structs.ACLToken{}
require.NoError(t, acl.Login(&structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method1.Name,
BearerToken: "fake-token",
},
Datacenter: "dc1",
}, &methodToken))
method2, err := upsertTestAuthMethod(codec, "root", "dc1", "")
require.NoError(t, err)
// try to update the token and change the method
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: methodToken.AccessorID,
SecretID: methodToken.SecretID,
AuthMethod: method2.Name,
Description: "updated token",
Local: true,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err = acl.TokenSet(&req, &resp)
requireErrorContains(t, err, "Cannot change AuthMethod")
})
t.Run("Update auth method linked token and let the SecretID and AuthMethod be defaulted", func(t *testing.T) {
acl := ACL{srv: s1}
testSessionID := testauth.StartSession()
defer testauth.ResetSession(testSessionID)
testauth.InstallSessionToken(testSessionID, "fake-token", "default", "demo", "abc123")
method, err := upsertTestAuthMethod(codec, "root", "dc1", testSessionID)
require.NoError(t, err)
_, err = upsertTestBindingRule(codec, "root", "dc1", method.Name, "", structs.BindingRuleBindTypeService, "demo")
require.NoError(t, err)
methodToken := structs.ACLToken{}
require.NoError(t, acl.Login(&structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-token",
},
Datacenter: "dc1",
}, &methodToken))
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: methodToken.AccessorID,
// SecretID: methodToken.SecretID,
// AuthMethod: method.Name,
Description: "updated token",
Local: true,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
require.NoError(t, acl.TokenSet(&req, &resp))
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
require.NotNil(t, token)
require.Len(t, token.Roles, 0)
require.Equal(t, "updated token", token.Description)
require.True(t, token.Local)
require.Equal(t, methodToken.SecretID, token.SecretID)
require.Equal(t, methodToken.AuthMethod, token.AuthMethod)
})
t.Run("Create it with invalid service identity (empty)", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{ServiceName: ""},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
requireErrorContains(t, err, "Service identity is missing the service name field")
})
t.Run("Create it with invalid service identity (too large)", func(t *testing.T) {
long := strings.Repeat("x", serviceIdentityNameMaxLength+1)
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{ServiceName: long},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.NotNil(t, err)
})
for _, test := range []struct {
name string
ok bool
}{
{"-abc", false},
{"abc-", false},
{"a-bc", true},
{"_abc", false},
{"abc_", false},
{"a_bc", true},
{":abc", false},
{"abc:", false},
{"a:bc", false},
{"Abc", false},
{"aBc", false},
{"abC", false},
{"0abc", true},
{"abc0", true},
{"a0bc", true},
} {
var testName string
if test.ok {
testName = "Create it with valid service identity (by regex): " + test.name
} else {
testName = "Create it with invalid service identity (by regex): " + test.name
}
t.Run(testName, func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{ServiceName: test.name},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
if test.ok {
require.NoError(t, err)
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
require.NotNil(t, token)
require.ElementsMatch(t, req.ACLToken.ServiceIdentities, token.ServiceIdentities)
} else {
require.NotNil(t, err)
}
})
}
t.Run("Create it with two of the same service identities", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{ServiceName: "example"},
&structs.ACLServiceIdentity{ServiceName: "example"},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.NoError(t, err)
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
require.NotNil(t, token)
require.Len(t, token.ServiceIdentities, 1)
})
t.Run("Create it with two of the same service identities and different DCs", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{
ServiceName: "example",
Datacenters: []string{"dc2", "dc3"},
},
&structs.ACLServiceIdentity{
ServiceName: "example",
Datacenters: []string{"dc1", "dc2"},
},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.NoError(t, err)
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
require.NotNil(t, token)
require.Len(t, token.ServiceIdentities, 1)
svcid := token.ServiceIdentities[0]
require.Equal(t, "example", svcid.ServiceName)
require.ElementsMatch(t, []string{"dc1", "dc2", "dc3"}, svcid.Datacenters)
})
t.Run("Create it with invalid service identity (datacenters set on local token)", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: true,
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{ServiceName: "foo", Datacenters: []string{"dc2"}},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
requireErrorContains(t, err, "cannot specify a list of datacenters on a local token")
})
for _, test := range []struct {
name string
offset time.Duration
errString string
errStringTTL string
}{
{"before create time", -5 * time.Minute, "ExpirationTime cannot be before CreateTime", ""},
{"too soon", 1 * time.Millisecond, "ExpirationTime cannot be less than", "ExpirationTime cannot be less than"},
{"too distant", 25 * time.Hour, "ExpirationTime cannot be more than", "ExpirationTime cannot be more than"},
} {
t.Run("Create it with an expiration time that is "+test.name, func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
ExpirationTime: timePointer(time.Now().Add(test.offset)),
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
if test.errString != "" {
requireErrorContains(t, err, test.errString)
} else {
require.NotNil(t, err)
}
})
t.Run("Create it with an expiration TTL that is "+test.name, func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
ExpirationTTL: test.offset,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
if test.errString != "" {
requireErrorContains(t, err, test.errStringTTL)
} else {
require.NotNil(t, err)
}
})
}
t.Run("Create it with expiration time AND expiration TTL set (error)", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
ExpirationTime: timePointer(time.Now().Add(4 * time.Second)),
ExpirationTTL: 4 * time.Second,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
requireErrorContains(t, err, "Expiration TTL and Expiration Time cannot both be set")
})
t.Run("Create it with expiration time using TTLs", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
ExpirationTTL: 4 * time.Second,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.NoError(t, err)
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
expectExpTime := resp.CreateTime.Add(4 * time.Second)
require.NotNil(t, token)
require.NotNil(t, token.AccessorID)
require.Equal(t, token.Description, "foobar")
require.Equal(t, token.AccessorID, resp.AccessorID)
requireTimeEquals(t, &expectExpTime, resp.ExpirationTime)
tokenID = token.AccessorID
})
var expTime time.Time
t.Run("Create it with expiration time", func(t *testing.T) {
expTime = time.Now().Add(4 * time.Second)
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "foobar",
Policies: nil,
Local: false,
ExpirationTime: &expTime,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.NoError(t, err)
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
require.NotNil(t, token)
require.NotNil(t, token.AccessorID)
require.Equal(t, token.Description, "foobar")
require.Equal(t, token.AccessorID, resp.AccessorID)
requireTimeEquals(t, &expTime, resp.ExpirationTime)
tokenID = token.AccessorID
})
// do not insert another test at this point: these tests need to be serial
t.Run("Update expiration time is not allowed", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "new-description",
AccessorID: tokenID,
ExpirationTime: timePointer(expTime.Add(-1 * time.Second)),
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
requireErrorContains(t, err, "Cannot change expiration time")
})
// do not insert another test at this point: these tests need to be serial
t.Run("Update anything except expiration time is ok - omit expiration time and let it default", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "new-description-1",
AccessorID: tokenID,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.NoError(t, err)
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
require.NotNil(t, token)
require.NotNil(t, token.AccessorID)
require.Equal(t, token.Description, "new-description-1")
require.Equal(t, token.AccessorID, resp.AccessorID)
requireTimeEquals(t, &expTime, resp.ExpirationTime)
})
t.Run("Update anything except expiration time is ok", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "new-description-2",
AccessorID: tokenID,
ExpirationTime: &expTime,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.NoError(t, err)
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
require.NotNil(t, token)
require.NotNil(t, token.AccessorID)
require.Equal(t, token.Description, "new-description-2")
require.Equal(t, token.AccessorID, resp.AccessorID)
requireTimeEquals(t, &expTime, resp.ExpirationTime)
})
t.Run("cannot update a token that is past its expiration time", func(t *testing.T) {
// create a token that will expire
expiringToken, err := upsertTestToken(codec, "root", "dc1", func(token *structs.ACLToken) {
token.ExpirationTTL = 11 * time.Millisecond
})
require.NoError(t, err)
time.Sleep(20 * time.Millisecond) // now 'expiringToken' is expired
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Description: "new-description",
AccessorID: expiringToken.AccessorID,
ExpirationTTL: 4 * time.Second,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err = acl.TokenSet(&req, &resp)
requireErrorContains(t, err, "Cannot find token")
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
2019-04-30 15:45:36 +00:00
func TestACLEndpoint_TokenSet_CustomID(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
// No Create Arg
t.Run("no create arg", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: "5d62a983-bcab-4e0c-9bcd-5dabebe3e273",
SecretID: "10a8ad77-2bdf-4939-a9d7-1b7de79d6beb",
Description: "foobar",
Policies: nil,
Local: false,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
// Use the Create Arg
t.Run("create arg", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: "5d62a983-bcab-4e0c-9bcd-5dabebe3e273",
SecretID: "10a8ad77-2bdf-4939-a9d7-1b7de79d6beb",
Description: "foobar",
Policies: nil,
Local: false,
},
Create: true,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.NoError(t, err)
// Get the token directly to validate that it exists
tokenResp, err := retrieveTestToken(codec, "root", "dc1", resp.AccessorID)
require.NoError(t, err)
token := tokenResp.Token
require.NotNil(t, token)
2019-04-30 15:45:36 +00:00
require.Equal(t, req.ACLToken.AccessorID, token.AccessorID)
require.Equal(t, req.ACLToken.SecretID, token.SecretID)
require.Equal(t, token.Description, "foobar")
})
// Reserved AccessorID
t.Run("reserved AccessorID", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: "00000000-0000-0000-0000-000000000073",
Description: "foobar",
Policies: nil,
Local: false,
},
Create: true,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
// Reserved SecretID
t.Run("reserved SecretID", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
SecretID: "00000000-0000-0000-0000-000000000073",
Description: "foobar",
Policies: nil,
Local: false,
},
Create: true,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
// Accessor is dup
t.Run("accessor Dup", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: "5d62a983-bcab-4e0c-9bcd-5dabebe3e273",
Description: "foobar",
Policies: nil,
Local: false,
},
Create: true,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
// Accessor is dup of secret
t.Run("accessor dup of secret", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: "10a8ad77-2bdf-4939-a9d7-1b7de79d6beb",
Description: "foobar",
Policies: nil,
Local: false,
},
Create: true,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
// Secret is dup of Accessor
t.Run("secret dup of accessor", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
SecretID: "5d62a983-bcab-4e0c-9bcd-5dabebe3e273",
Description: "foobar",
Policies: nil,
Local: false,
},
Create: true,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
// Secret is dup
t.Run("secret dup", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
SecretID: "10a8ad77-2bdf-4939-a9d7-1b7de79d6beb",
Description: "foobar",
Policies: nil,
Local: false,
},
Create: true,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
// Update Accessor attempt
t.Run("update accessor", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: "75a0d6a9-6882-4f7a-a053-906db1d55a73",
SecretID: "10a8ad77-2bdf-4939-a9d7-1b7de79d6beb",
Description: "foobar",
Policies: nil,
Local: false,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
// Update Accessor attempt - with Create
t.Run("update accessor create", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: "75a0d6a9-6882-4f7a-a053-906db1d55a73",
SecretID: "10a8ad77-2bdf-4939-a9d7-1b7de79d6beb",
Description: "foobar",
Policies: nil,
Local: false,
},
Create: true,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
// Update Secret attempt
t.Run("update secret", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: "5d62a983-bcab-4e0c-9bcd-5dabebe3e273",
SecretID: "f551f807-b3a7-4483-9ade-97230c974bf3",
Description: "foobar",
Policies: nil,
Local: false,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
// Update Secret attempt - with Create
t.Run("update secret create", func(t *testing.T) {
req := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: "5d62a983-bcab-4e0c-9bcd-5dabebe3e273",
SecretID: "f551f807-b3a7-4483-9ade-97230c974bf3",
Description: "foobar",
Policies: nil,
Local: false,
},
Create: true,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLToken{}
err := acl.TokenSet(&req, &resp)
require.Error(t, err)
})
}
func TestACLEndpoint_TokenSet_anon(t *testing.T) {
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
policy, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
acl := ACL{srv: s1}
// Assign the policies to a token
tokenUpsertReq := structs.ACLTokenSetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
ACLToken: structs.ACLToken{
AccessorID: structs.ACLTokenAnonymousID,
Policies: []structs.ACLTokenPolicyLink{
structs.ACLTokenPolicyLink{
ID: policy.ID,
},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
token := structs.ACLToken{}
err = acl.TokenSet(&tokenUpsertReq, &token)
require.NoError(t, err)
require.NotEmpty(t, token.SecretID)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
tokenResp, err := retrieveTestToken(codec, "root", "dc1", structs.ACLTokenAnonymousID)
require.Equal(t, len(tokenResp.Token.Policies), 1)
require.Equal(t, tokenResp.Token.Policies[0].ID, policy.ID)
2019-04-30 15:45:36 +00:00
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_TokenDelete(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
dir2, s2 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.Datacenter = "dc2"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
// token replication is required to test deleting non-local tokens in secondary dc
c.ACLTokenReplication = true
})
defer os.RemoveAll(dir2)
defer s2.Shutdown()
codec2 := rpcClient(t, s2)
defer codec2.Close()
ACL Token Persistence and Reloading (#5328) This PR adds two features which will be useful for operators when ACLs are in use. 1. Tokens set in configuration files are now reloadable. 2. If `acl.enable_token_persistence` is set to `true` in the configuration, tokens set via the `v1/agent/token` endpoint are now persisted to disk and loaded when the agent starts (or during configuration reload) Note that token persistence is opt-in so our users who do not want tokens on the local disk will see no change. Some other secondary changes: * Refactored a bunch of places where the replication token is retrieved from the token store. This token isn't just for replicating ACLs and now it is named accordingly. * Allowed better paths in the `v1/agent/token/` API. Instead of paths like: `v1/agent/token/acl_replication_token` the path can now be just `v1/agent/token/replication`. The old paths remain to be valid. * Added a couple new API functions to set tokens via the new paths. Deprecated the old ones and pointed to the new names. The names are also generally better and don't imply that what you are setting is for ACLs but rather are setting ACL tokens. There is a minor semantic difference there especially for the replication token as again, its no longer used only for ACL token/policy replication. The new functions will detect 404s and fallback to using the older token paths when talking to pre-1.4.3 agents. * Docs updated to reflect the API additions and to show using the new endpoints. * Updated the ACL CLI set-agent-tokens command to use the non-deprecated APIs.
2019-02-27 19:28:31 +00:00
s2.tokens.UpdateReplicationToken("root", tokenStore.TokenSourceConfig)
testrpc.WaitForLeader(t, s1.RPC, "dc1")
testrpc.WaitForLeader(t, s2.RPC, "dc2")
// Try to join
joinWAN(t, s2, s1)
waitForNewACLs(t, s1)
waitForNewACLs(t, s2)
// Ensure s2 is authoritative.
waitForNewACLReplication(t, s2, structs.ACLReplicateTokens, 1, 1, 0)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
acl := ACL{srv: s1}
acl2 := ACL{srv: s2}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
existingToken, err := upsertTestToken(codec, "root", "dc1", nil)
require.NoError(t, err)
t.Run("deletes a token that has an expiration time in the future", func(t *testing.T) {
// create a token that will expire
testToken, err := upsertTestToken(codec, "root", "dc1", func(token *structs.ACLToken) {
token.ExpirationTTL = 4 * time.Second
})
require.NoError(t, err)
// Make sure the token is listable
tokenResp, err := retrieveTestToken(codec, "root", "dc1", testToken.AccessorID)
require.NoError(t, err)
require.NotNil(t, tokenResp.Token)
// Now try to delete it (this should work).
req := structs.ACLTokenDeleteRequest{
Datacenter: "dc1",
TokenID: testToken.AccessorID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var resp string
err = acl.TokenDelete(&req, &resp)
require.NoError(t, err)
// Make sure the token is gone
tokenResp, err = retrieveTestToken(codec, "root", "dc1", testToken.AccessorID)
require.NoError(t, err)
require.Nil(t, tokenResp.Token)
})
t.Run("deletes a token that is past its expiration time", func(t *testing.T) {
// create a token that will expire
expiringToken, err := upsertTestToken(codec, "root", "dc1", func(token *structs.ACLToken) {
token.ExpirationTTL = 11 * time.Millisecond
})
require.NoError(t, err)
time.Sleep(20 * time.Millisecond) // now 'expiringToken' is expired
// Make sure the token is not listable (filtered due to expiry)
tokenResp, err := retrieveTestToken(codec, "root", "dc1", expiringToken.AccessorID)
require.NoError(t, err)
require.Nil(t, tokenResp.Token)
// Now try to delete it (this should work).
req := structs.ACLTokenDeleteRequest{
Datacenter: "dc1",
TokenID: expiringToken.AccessorID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var resp string
err = acl.TokenDelete(&req, &resp)
require.NoError(t, err)
// Make sure the token is still gone (this time it's actually gone)
tokenResp, err = retrieveTestToken(codec, "root", "dc1", expiringToken.AccessorID)
require.NoError(t, err)
require.Nil(t, tokenResp.Token)
})
t.Run("deletes a token", func(t *testing.T) {
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
req := structs.ACLTokenDeleteRequest{
Datacenter: "dc1",
TokenID: existingToken.AccessorID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var resp string
err = acl.TokenDelete(&req, &resp)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// Make sure the token is gone
tokenResp, err := retrieveTestToken(codec, "root", "dc1", existingToken.AccessorID)
require.Nil(t, tokenResp.Token)
require.NoError(t, err)
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Run("can't delete itself", func(t *testing.T) {
readReq := structs.ACLTokenGetRequest{
Datacenter: "dc1",
TokenID: "root",
TokenIDType: structs.ACLTokenSecret,
QueryOptions: structs.QueryOptions{Token: "root"},
}
var out structs.ACLTokenResponse
err := acl.TokenRead(&readReq, &out)
require.NoError(t, err)
req := structs.ACLTokenDeleteRequest{
Datacenter: "dc1",
TokenID: out.Token.AccessorID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var resp string
err = acl.TokenDelete(&req, &resp)
require.EqualError(t, err, "Deletion of the request's authorization token is not permitted")
})
t.Run("errors when token doesn't exist", func(t *testing.T) {
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
fakeID, err := uuid.GenerateUUID()
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
req := structs.ACLTokenDeleteRequest{
Datacenter: "dc1",
TokenID: fakeID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var resp string
err = acl.TokenDelete(&req, &resp)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// token should be nil
tokenResp, err := retrieveTestToken(codec, "root", "dc1", existingToken.AccessorID)
require.Nil(t, tokenResp.Token)
require.NoError(t, err)
})
2019-03-06 19:47:33 +00:00
t.Run("don't segfault when attempting to delete non existent token in secondary dc", func(t *testing.T) {
fakeID, err := uuid.GenerateUUID()
require.NoError(t, err)
req := structs.ACLTokenDeleteRequest{
Datacenter: "dc2",
TokenID: fakeID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var resp string
err = acl2.TokenDelete(&req, &resp)
require.NoError(t, err)
// token should be nil
tokenResp, err := retrieveTestToken(codec2, "root", "dc1", existingToken.AccessorID)
require.Nil(t, tokenResp.Token)
require.NoError(t, err)
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
func TestACLEndpoint_TokenDelete_anon(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
req := structs.ACLTokenDeleteRequest{
Datacenter: "dc1",
TokenID: structs.ACLTokenAnonymousID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var resp string
err := acl.TokenDelete(&req, &resp)
require.EqualError(t, err, "Delete operation not permitted on the anonymous token")
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// Make sure the token is still there
tokenResp, err := retrieveTestToken(codec, "root", "dc1", structs.ACLTokenAnonymousID)
require.NotNil(t, tokenResp.Token)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_TokenList(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
t1, err := upsertTestToken(codec, "root", "dc1", nil)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t2, err := upsertTestToken(codec, "root", "dc1", nil)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
masterTokenAccessorID, err := retrieveTestTokenAccessorForSecret(codec, "root", "dc1", "root")
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Run("normal", func(t *testing.T) {
2019-11-01 20:11:44 +00:00
// this will still be racey even with inserting the token + ttl inside the test function
// however previously inserting it outside of the subtest func resulted in this being
// extra flakey due to there being more code that needed to run to setup the subtest
// between when we inserted the token and when we performed the listing.
t3, err := upsertTestToken(codec, "root", "dc1", func(token *structs.ACLToken) {
token.ExpirationTTL = 50 * time.Millisecond
})
require.NoError(t, err)
req := structs.ACLTokenListRequest{
Datacenter: "dc1",
QueryOptions: structs.QueryOptions{Token: "root"},
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
resp := structs.ACLTokenListResponse{}
err = acl.TokenList(&req, &resp)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
tokens := []string{
masterTokenAccessorID,
structs.ACLTokenAnonymousID,
t1.AccessorID,
t2.AccessorID,
t3.AccessorID,
}
require.ElementsMatch(t, gatherIDs(t, resp.Tokens), tokens)
})
2019-11-01 20:11:44 +00:00
time.Sleep(50 * time.Millisecond) // now 't3' is expired
t.Run("filter expired", func(t *testing.T) {
req := structs.ACLTokenListRequest{
Datacenter: "dc1",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLTokenListResponse{}
err = acl.TokenList(&req, &resp)
require.NoError(t, err)
tokens := []string{
masterTokenAccessorID,
structs.ACLTokenAnonymousID,
t1.AccessorID,
t2.AccessorID,
}
require.ElementsMatch(t, gatherIDs(t, resp.Tokens), tokens)
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_TokenBatchRead(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
t1, err := upsertTestToken(codec, "root", "dc1", nil)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t2, err := upsertTestToken(codec, "root", "dc1", nil)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t3, err := upsertTestToken(codec, "root", "dc1", func(token *structs.ACLToken) {
token.ExpirationTTL = 4 * time.Second
})
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Run("normal", func(t *testing.T) {
tokens := []string{t1.AccessorID, t2.AccessorID, t3.AccessorID}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
req := structs.ACLTokenBatchGetRequest{
Datacenter: "dc1",
AccessorIDs: tokens,
QueryOptions: structs.QueryOptions{Token: "root"},
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
resp := structs.ACLTokenBatchResponse{}
err = acl.TokenBatchRead(&req, &resp)
require.NoError(t, err)
require.ElementsMatch(t, gatherIDs(t, resp.Tokens), tokens)
})
time.Sleep(20 * time.Millisecond) // now 't3' is expired
t.Run("returns expired tokens", func(t *testing.T) {
tokens := []string{t1.AccessorID, t2.AccessorID, t3.AccessorID}
req := structs.ACLTokenBatchGetRequest{
Datacenter: "dc1",
AccessorIDs: tokens,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLTokenBatchResponse{}
err = acl.TokenBatchRead(&req, &resp)
require.NoError(t, err)
require.ElementsMatch(t, gatherIDs(t, resp.Tokens), tokens)
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_PolicyRead(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
policy, err := upsertTestPolicy(codec, "root", "dc1")
if err != nil {
t.Fatalf("err: %v", err)
}
acl := ACL{srv: s1}
req := structs.ACLPolicyGetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
PolicyID: policy.ID,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLPolicyResponse{}
err = acl.PolicyRead(&req, &resp)
if err != nil {
t.Fatalf("err: %v", err)
}
if !reflect.DeepEqual(resp.Policy, policy) {
t.Fatalf("tokens are not equal: %v != %v", resp.Policy, policy)
}
}
func TestACLEndpoint_PolicyBatchRead(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
p1, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
p2, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
acl := ACL{srv: s1}
policies := []string{p1.ID, p2.ID}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
req := structs.ACLPolicyBatchGetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
PolicyIDs: policies,
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLPolicyBatchResponse{}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
err = acl.PolicyBatchRead(&req, &resp)
require.NoError(t, err)
require.ElementsMatch(t, gatherIDs(t, resp.Policies), []string{p1.ID, p2.ID})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_PolicySet(t *testing.T) {
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
var policyID string
t.Run("Create it", func(t *testing.T) {
req := structs.ACLPolicySetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
Policy: structs.ACLPolicy{
Description: "foobar",
Name: "baz",
Rules: "service \"\" { policy = \"read\" }",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLPolicy{}
err := acl.PolicySet(&req, &resp)
require.NoError(t, err)
require.NotNil(t, resp.ID)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// Get the policy directly to validate that it exists
policyResp, err := retrieveTestPolicy(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
policy := policyResp.Policy
require.NotNil(t, policy.ID)
require.Equal(t, policy.Description, "foobar")
require.Equal(t, policy.Name, "baz")
require.Equal(t, policy.Rules, "service \"\" { policy = \"read\" }")
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
policyID = policy.ID
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
2019-04-30 15:45:36 +00:00
t.Run("Name Dup", func(t *testing.T) {
req := structs.ACLPolicySetRequest{
Datacenter: "dc1",
Policy: structs.ACLPolicy{
Description: "foobar",
Name: "baz",
Rules: "service \"\" { policy = \"read\" }",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLPolicy{}
err := acl.PolicySet(&req, &resp)
require.Error(t, err)
})
t.Run("Update it", func(t *testing.T) {
req := structs.ACLPolicySetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
Policy: structs.ACLPolicy{
ID: policyID,
Description: "bat",
Name: "bar",
Rules: "service \"\" { policy = \"write\" }",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLPolicy{}
err := acl.PolicySet(&req, &resp)
require.NoError(t, err)
require.NotNil(t, resp.ID)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// Get the policy directly to validate that it exists
policyResp, err := retrieveTestPolicy(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
policy := policyResp.Policy
require.NotNil(t, policy.ID)
require.Equal(t, policy.Description, "bat")
require.Equal(t, policy.Name, "bar")
require.Equal(t, policy.Rules, "service \"\" { policy = \"write\" }")
})
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
2019-04-30 15:45:36 +00:00
func TestACLEndpoint_PolicySet_CustomID(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
// Attempt to create policy with ID
req := structs.ACLPolicySetRequest{
Datacenter: "dc1",
Policy: structs.ACLPolicy{
ID: "7ee166a5-b4b7-453c-bdc0-bca8ce50823e",
Description: "foobar",
Name: "baz",
Rules: "service \"\" { policy = \"read\" }",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLPolicy{}
err := acl.PolicySet(&req, &resp)
require.Error(t, err)
}
func TestACLEndpoint_PolicySet_globalManagement(t *testing.T) {
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
// Can't change the rules
{
req := structs.ACLPolicySetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
Policy: structs.ACLPolicy{
ID: structs.ACLPolicyGlobalManagementID,
Name: "foobar", // This is required to get past validation
Rules: "service \"\" { policy = \"write\" }",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLPolicy{}
err := acl.PolicySet(&req, &resp)
require.EqualError(t, err, "Changing the Rules for the builtin global-management policy is not permitted")
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
// Can rename it
{
req := structs.ACLPolicySetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
Policy: structs.ACLPolicy{
ID: structs.ACLPolicyGlobalManagementID,
Name: "foobar",
Rules: structs.ACLPolicyGlobalManagement,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLPolicy{}
err := acl.PolicySet(&req, &resp)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// Get the policy again
policyResp, err := retrieveTestPolicy(codec, "root", "dc1", structs.ACLPolicyGlobalManagementID)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
policy := policyResp.Policy
require.Equal(t, policy.ID, structs.ACLPolicyGlobalManagementID)
require.Equal(t, policy.Name, "foobar")
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
}
func TestACLEndpoint_PolicyDelete(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
existingPolicy, err := upsertTestPolicy(codec, "root", "dc1")
if err != nil {
t.Fatalf("err: %v", err)
}
acl := ACL{srv: s1}
req := structs.ACLPolicyDeleteRequest{
Datacenter: "dc1",
PolicyID: existingPolicy.ID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var resp string
err = acl.PolicyDelete(&req, &resp)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// Make sure the policy is gone
tokenResp, err := retrieveTestPolicy(codec, "root", "dc1", existingPolicy.ID)
require.Nil(t, tokenResp.Policy)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_PolicyDelete_globalManagement(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
req := structs.ACLPolicyDeleteRequest{
Datacenter: "dc1",
PolicyID: structs.ACLPolicyGlobalManagementID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var resp string
err := acl.PolicyDelete(&req, &resp)
require.EqualError(t, err, "Delete operation not permitted on the builtin global-management policy")
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_PolicyList(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
p1, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
p2, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
acl := ACL{srv: s1}
req := structs.ACLPolicyListRequest{
Datacenter: "dc1",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLPolicyListResponse{}
err = acl.PolicyList(&req, &resp)
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
policies := []string{
structs.ACLPolicyGlobalManagementID,
p1.ID,
p2.ID,
}
require.ElementsMatch(t, gatherIDs(t, resp.Policies), policies)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_PolicyResolve(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
p1, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
p2, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
acl := ACL{srv: s1}
policies := []string{p1.ID, p2.ID}
// Assign the policies to a token
tokenUpsertReq := structs.ACLTokenSetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Policies: []structs.ACLTokenPolicyLink{
structs.ACLTokenPolicyLink{
ID: p1.ID,
},
structs.ACLTokenPolicyLink{
ID: p2.ID,
},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
token := structs.ACLToken{}
err = acl.TokenSet(&tokenUpsertReq, &token)
require.NoError(t, err)
require.NotEmpty(t, token.SecretID)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
resp := structs.ACLPolicyBatchResponse{}
req := structs.ACLPolicyBatchGetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: "dc1",
PolicyIDs: []string{p1.ID, p2.ID},
QueryOptions: structs.QueryOptions{Token: token.SecretID},
}
err = acl.PolicyResolve(&req, &resp)
require.NoError(t, err)
require.ElementsMatch(t, gatherIDs(t, resp.Policies), policies)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
}
func TestACLEndpoint_RoleRead(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
role, err := upsertTestRole(codec, "root", "dc1")
require.NoError(t, err)
acl := ACL{srv: s1}
req := structs.ACLRoleGetRequest{
Datacenter: "dc1",
RoleID: role.ID,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLRoleResponse{}
err = acl.RoleRead(&req, &resp)
require.NoError(t, err)
require.Equal(t, role, resp.Role)
}
func TestACLEndpoint_RoleBatchRead(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
r1, err := upsertTestRole(codec, "root", "dc1")
require.NoError(t, err)
r2, err := upsertTestRole(codec, "root", "dc1")
require.NoError(t, err)
acl := ACL{srv: s1}
roles := []string{r1.ID, r2.ID}
req := structs.ACLRoleBatchGetRequest{
Datacenter: "dc1",
RoleIDs: roles,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLRoleBatchResponse{}
err = acl.RoleBatchRead(&req, &resp)
require.NoError(t, err)
require.ElementsMatch(t, gatherIDs(t, resp.Roles), roles)
}
func TestACLEndpoint_RoleSet(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
var roleID string
testPolicy1, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
testPolicy2, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
t.Run("Create it", func(t *testing.T) {
req := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
Description: "foobar",
Name: "baz",
Policies: []structs.ACLRolePolicyLink{
structs.ACLRolePolicyLink{
ID: testPolicy1.ID,
},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLRole{}
err := acl.RoleSet(&req, &resp)
require.NoError(t, err)
require.NotNil(t, resp.ID)
// Get the role directly to validate that it exists
roleResp, err := retrieveTestRole(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
role := roleResp.Role
require.NotNil(t, role.ID)
require.Equal(t, role.Description, "foobar")
require.Equal(t, role.Name, "baz")
require.Len(t, role.Policies, 1)
require.Equal(t, testPolicy1.ID, role.Policies[0].ID)
roleID = role.ID
})
t.Run("Update it", func(t *testing.T) {
req := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
ID: roleID,
Description: "bat",
Name: "bar",
Policies: []structs.ACLRolePolicyLink{
structs.ACLRolePolicyLink{
ID: testPolicy2.ID,
},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLRole{}
err := acl.RoleSet(&req, &resp)
require.NoError(t, err)
require.NotNil(t, resp.ID)
// Get the role directly to validate that it exists
roleResp, err := retrieveTestRole(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
role := roleResp.Role
require.NotNil(t, role.ID)
require.Equal(t, role.Description, "bat")
require.Equal(t, role.Name, "bar")
require.Len(t, role.Policies, 1)
require.Equal(t, testPolicy2.ID, role.Policies[0].ID)
})
t.Run("Create it using Policies linked by id and name", func(t *testing.T) {
policy1, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
policy2, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
req := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
Description: "foobar",
Name: "baz",
Policies: []structs.ACLRolePolicyLink{
structs.ACLRolePolicyLink{
ID: policy1.ID,
},
structs.ACLRolePolicyLink{
Name: policy2.Name,
},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLRole{}
err = acl.RoleSet(&req, &resp)
require.NoError(t, err)
require.NotNil(t, resp.ID)
// Delete both policies to ensure that we skip resolving ID->Name
// in the returned data.
require.NoError(t, deleteTestPolicy(codec, "root", "dc1", policy1.ID))
require.NoError(t, deleteTestPolicy(codec, "root", "dc1", policy2.ID))
// Get the role directly to validate that it exists
roleResp, err := retrieveTestRole(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
role := roleResp.Role
require.NotNil(t, role.ID)
require.Equal(t, role.Description, "foobar")
require.Equal(t, role.Name, "baz")
require.Len(t, role.Policies, 0)
})
roleNameGen := func(t *testing.T) string {
t.Helper()
name, err := uuid.GenerateUUID()
require.NoError(t, err)
return name
}
t.Run("Create it with invalid service identity (empty)", func(t *testing.T) {
req := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
Description: "foobar",
Name: roleNameGen(t),
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{ServiceName: ""},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLRole{}
err := acl.RoleSet(&req, &resp)
requireErrorContains(t, err, "Service identity is missing the service name field")
})
t.Run("Create it with invalid service identity (too large)", func(t *testing.T) {
long := strings.Repeat("x", serviceIdentityNameMaxLength+1)
req := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
Description: "foobar",
Name: roleNameGen(t),
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{ServiceName: long},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLRole{}
err := acl.RoleSet(&req, &resp)
require.NotNil(t, err)
})
for _, test := range []struct {
name string
ok bool
}{
{"-abc", false},
{"abc-", false},
{"a-bc", true},
{"_abc", false},
{"abc_", false},
{"a_bc", true},
{":abc", false},
{"abc:", false},
{"a:bc", false},
{"Abc", false},
{"aBc", false},
{"abC", false},
{"0abc", true},
{"abc0", true},
{"a0bc", true},
} {
var testName string
if test.ok {
testName = "Create it with valid service identity (by regex): " + test.name
} else {
testName = "Create it with invalid service identity (by regex): " + test.name
}
t.Run(testName, func(t *testing.T) {
req := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
Description: "foobar",
Name: roleNameGen(t),
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{ServiceName: test.name},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLRole{}
err := acl.RoleSet(&req, &resp)
if test.ok {
require.NoError(t, err)
// Get the token directly to validate that it exists
roleResp, err := retrieveTestRole(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
role := roleResp.Role
require.ElementsMatch(t, req.Role.ServiceIdentities, role.ServiceIdentities)
} else {
require.NotNil(t, err)
}
})
}
t.Run("Create it with two of the same service identities", func(t *testing.T) {
req := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
Description: "foobar",
Name: roleNameGen(t),
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{ServiceName: "example"},
&structs.ACLServiceIdentity{ServiceName: "example"},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLRole{}
err := acl.RoleSet(&req, &resp)
require.NoError(t, err)
// Get the role directly to validate that it exists
roleResp, err := retrieveTestRole(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
role := roleResp.Role
require.Len(t, role.ServiceIdentities, 1)
})
t.Run("Create it with two of the same service identities and different DCs", func(t *testing.T) {
req := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
Description: "foobar",
Name: roleNameGen(t),
ServiceIdentities: []*structs.ACLServiceIdentity{
&structs.ACLServiceIdentity{
ServiceName: "example",
Datacenters: []string{"dc2", "dc3"},
},
&structs.ACLServiceIdentity{
ServiceName: "example",
Datacenters: []string{"dc1", "dc2"},
},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLRole{}
err := acl.RoleSet(&req, &resp)
require.NoError(t, err)
// Get the role directly to validate that it exists
roleResp, err := retrieveTestRole(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
role := roleResp.Role
require.Len(t, role.ServiceIdentities, 1)
svcid := role.ServiceIdentities[0]
require.Equal(t, "example", svcid.ServiceName)
require.ElementsMatch(t, []string{"dc1", "dc2", "dc3"}, svcid.Datacenters)
})
}
func TestACLEndpoint_RoleSet_names(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
testPolicy1, err := upsertTestPolicy(codec, "root", "dc1")
require.NoError(t, err)
for _, test := range []struct {
name string
ok bool
}{
{"", false},
{"-bad", true},
{"bad-", true},
{"bad?bad", false},
{strings.Repeat("x", 257), false},
{strings.Repeat("x", 256), true},
{"-abc", true},
{"abc-", true},
{"a-bc", true},
{"_abc", true},
{"abc_", true},
{"a_bc", true},
{":abc", false},
{"abc:", false},
{"a:bc", false},
{"Abc", true},
{"aBc", true},
{"abC", true},
{"0abc", true},
{"abc0", true},
{"a0bc", true},
} {
var testName string
if test.ok {
testName = "create with valid name: " + test.name
} else {
testName = "create with invalid name: " + test.name
}
t.Run(testName, func(t *testing.T) {
// cleanup from a prior insertion that may have succeeded
require.NoError(t, deleteTestRoleByName(codec, "root", "dc1", test.name))
req := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
Name: test.name,
Description: "foobar",
Policies: []structs.ACLRolePolicyLink{
structs.ACLRolePolicyLink{
ID: testPolicy1.ID,
},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLRole{}
err := acl.RoleSet(&req, &resp)
if test.ok {
require.NoError(t, err)
roleResp, err := retrieveTestRole(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
role := roleResp.Role
require.Equal(t, test.name, role.Name)
} else {
require.Error(t, err)
}
})
}
}
func TestACLEndpoint_RoleDelete(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
existingRole, err := upsertTestRole(codec, "root", "dc1")
require.NoError(t, err)
acl := ACL{srv: s1}
req := structs.ACLRoleDeleteRequest{
Datacenter: "dc1",
RoleID: existingRole.ID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var resp string
err = acl.RoleDelete(&req, &resp)
require.NoError(t, err)
// Make sure the role is gone
roleResp, err := retrieveTestRole(codec, "root", "dc1", existingRole.ID)
require.Nil(t, roleResp.Role)
}
func TestACLEndpoint_RoleList(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
r1, err := upsertTestRole(codec, "root", "dc1")
require.NoError(t, err)
r2, err := upsertTestRole(codec, "root", "dc1")
require.NoError(t, err)
acl := ACL{srv: s1}
req := structs.ACLRoleListRequest{
Datacenter: "dc1",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLRoleListResponse{}
err = acl.RoleList(&req, &resp)
require.NoError(t, err)
require.ElementsMatch(t, gatherIDs(t, resp.Roles), []string{r1.ID, r2.ID})
}
func TestACLEndpoint_RoleResolve(t *testing.T) {
t.Parallel()
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
testrpc.WaitForLeader(t, s1.RPC, "dc1")
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
t.Run("Normal", func(t *testing.T) {
r1, err := upsertTestRole(codec, "root", "dc1")
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
r2, err := upsertTestRole(codec, "root", "dc1")
require.NoError(t, err)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
acl := ACL{srv: s1}
// Assign the roles to a token
tokenUpsertReq := structs.ACLTokenSetRequest{
Datacenter: "dc1",
ACLToken: structs.ACLToken{
Roles: []structs.ACLTokenRoleLink{
structs.ACLTokenRoleLink{
ID: r1.ID,
},
structs.ACLTokenRoleLink{
ID: r2.ID,
},
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
token := structs.ACLToken{}
err = acl.TokenSet(&tokenUpsertReq, &token)
require.NoError(t, err)
require.NotEmpty(t, token.SecretID)
resp := structs.ACLRoleBatchResponse{}
req := structs.ACLRoleBatchGetRequest{
Datacenter: "dc1",
RoleIDs: []string{r1.ID, r2.ID},
QueryOptions: structs.QueryOptions{Token: token.SecretID},
}
err = acl.RoleResolve(&req, &resp)
require.NoError(t, err)
require.ElementsMatch(t, gatherIDs(t, resp.Roles), []string{r1.ID, r2.ID})
})
}
func TestACLEndpoint_AuthMethodSet(t *testing.T) {
t.Parallel()
tempDir, err := ioutil.TempDir("", "consul")
require.NoError(t, err)
defer os.RemoveAll(tempDir)
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
newAuthMethod := func(name string) structs.ACLAuthMethod {
return structs.ACLAuthMethod{
Name: name,
Description: "test",
Type: "testing",
}
}
t.Run("Create", func(t *testing.T) {
reqMethod := newAuthMethod("test")
req := structs.ACLAuthMethodSetRequest{
Datacenter: "dc1",
AuthMethod: reqMethod,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLAuthMethod{}
err := acl.AuthMethodSet(&req, &resp)
require.NoError(t, err)
// Get the method directly to validate that it exists
methodResp, err := retrieveTestAuthMethod(codec, "root", "dc1", resp.Name)
require.NoError(t, err)
method := methodResp.AuthMethod
require.Equal(t, method.Name, "test")
require.Equal(t, method.Description, "test")
require.Equal(t, method.Type, "testing")
})
t.Run("Update fails; not allowed to change types", func(t *testing.T) {
reqMethod := newAuthMethod("test")
reqMethod.Type = "invalid"
req := structs.ACLAuthMethodSetRequest{
Datacenter: "dc1",
AuthMethod: reqMethod,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLAuthMethod{}
err := acl.AuthMethodSet(&req, &resp)
require.Error(t, err)
})
t.Run("Update - allow type to default", func(t *testing.T) {
reqMethod := newAuthMethod("test")
reqMethod.Description = "test modified 1"
reqMethod.Type = "" // unset
req := structs.ACLAuthMethodSetRequest{
Datacenter: "dc1",
AuthMethod: reqMethod,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLAuthMethod{}
err := acl.AuthMethodSet(&req, &resp)
require.NoError(t, err)
// Get the method directly to validate that it exists
methodResp, err := retrieveTestAuthMethod(codec, "root", "dc1", resp.Name)
require.NoError(t, err)
method := methodResp.AuthMethod
require.Equal(t, method.Name, "test")
require.Equal(t, method.Description, "test modified 1")
require.Equal(t, method.Type, "testing")
})
t.Run("Update - specify type", func(t *testing.T) {
reqMethod := newAuthMethod("test")
reqMethod.Description = "test modified 2"
req := structs.ACLAuthMethodSetRequest{
Datacenter: "dc1",
AuthMethod: reqMethod,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLAuthMethod{}
err := acl.AuthMethodSet(&req, &resp)
require.NoError(t, err)
// Get the method directly to validate that it exists
methodResp, err := retrieveTestAuthMethod(codec, "root", "dc1", resp.Name)
require.NoError(t, err)
method := methodResp.AuthMethod
require.Equal(t, method.Name, "test")
require.Equal(t, method.Description, "test modified 2")
require.Equal(t, method.Type, "testing")
})
t.Run("Create with no name", func(t *testing.T) {
req := structs.ACLAuthMethodSetRequest{
Datacenter: "dc1",
AuthMethod: newAuthMethod(""),
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLAuthMethod{}
err := acl.AuthMethodSet(&req, &resp)
require.Error(t, err)
})
t.Run("Create with invalid type", func(t *testing.T) {
req := structs.ACLAuthMethodSetRequest{
Datacenter: "dc1",
AuthMethod: structs.ACLAuthMethod{
Name: "invalid",
Description: "invalid test",
Type: "invalid",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLAuthMethod{}
err := acl.AuthMethodSet(&req, &resp)
require.Error(t, err)
})
for _, test := range []struct {
name string
ok bool
}{
{strings.Repeat("x", 129), false},
{strings.Repeat("x", 128), true},
{"-abc", true},
{"abc-", true},
{"a-bc", true},
{"_abc", true},
{"abc_", true},
{"a_bc", true},
{":abc", false},
{"abc:", false},
{"a:bc", false},
{"Abc", true},
{"aBc", true},
{"abC", true},
{"0abc", true},
{"abc0", true},
{"a0bc", true},
} {
var testName string
if test.ok {
testName = "Create with valid name (by regex): " + test.name
} else {
testName = "Create with invalid name (by regex): " + test.name
}
t.Run(testName, func(t *testing.T) {
req := structs.ACLAuthMethodSetRequest{
Datacenter: "dc1",
AuthMethod: newAuthMethod(test.name),
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLAuthMethod{}
err := acl.AuthMethodSet(&req, &resp)
if test.ok {
require.NoError(t, err)
// Get the method directly to validate that it exists
methodResp, err := retrieveTestAuthMethod(codec, "root", "dc1", resp.Name)
require.NoError(t, err)
method := methodResp.AuthMethod
require.Equal(t, method.Name, test.name)
require.Equal(t, method.Type, "testing")
} else {
require.Error(t, err)
}
})
}
}
func TestACLEndpoint_AuthMethodDelete(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
testSessionID := testauth.StartSession()
defer testauth.ResetSession(testSessionID)
existingMethod, err := upsertTestAuthMethod(codec, "root", "dc1", testSessionID)
require.NoError(t, err)
acl := ACL{srv: s1}
t.Run("normal", func(t *testing.T) {
req := structs.ACLAuthMethodDeleteRequest{
Datacenter: "dc1",
AuthMethodName: existingMethod.Name,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var ignored bool
err = acl.AuthMethodDelete(&req, &ignored)
require.NoError(t, err)
// Make sure the method is gone
methodResp, err := retrieveTestAuthMethod(codec, "root", "dc1", existingMethod.Name)
require.NoError(t, err)
require.Nil(t, methodResp.AuthMethod)
})
t.Run("delete something that doesn't exist", func(t *testing.T) {
req := structs.ACLAuthMethodDeleteRequest{
Datacenter: "dc1",
AuthMethodName: "missing",
WriteRequest: structs.WriteRequest{Token: "root"},
}
var ignored bool
err = acl.AuthMethodDelete(&req, &ignored)
require.NoError(t, err)
})
}
// Deleting an auth method atomically deletes all rules and tokens as well.
func TestACLEndpoint_AuthMethodDelete_RuleAndTokenCascade(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
testSessionID1 := testauth.StartSession()
defer testauth.ResetSession(testSessionID1)
testauth.InstallSessionToken(testSessionID1, "fake-token1", "default", "abc", "abc123")
testSessionID2 := testauth.StartSession()
defer testauth.ResetSession(testSessionID2)
testauth.InstallSessionToken(testSessionID2, "fake-token2", "default", "abc", "abc123")
createToken := func(methodName, bearerToken string) *structs.ACLToken {
acl := ACL{srv: s1}
resp := structs.ACLToken{}
require.NoError(t, acl.Login(&structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: methodName,
BearerToken: bearerToken,
},
Datacenter: "dc1",
}, &resp))
return &resp
}
method1, err := upsertTestAuthMethod(codec, "root", "dc1", testSessionID1)
require.NoError(t, err)
i1_r1, err := upsertTestBindingRule(
codec, "root", "dc1",
method1.Name,
"serviceaccount.name==abc",
structs.BindingRuleBindTypeService,
"abc",
)
require.NoError(t, err)
i1_r2, err := upsertTestBindingRule(
codec, "root", "dc1",
method1.Name,
"serviceaccount.name==def",
structs.BindingRuleBindTypeService,
"def",
)
require.NoError(t, err)
i1_t1 := createToken(method1.Name, "fake-token1")
i1_t2 := createToken(method1.Name, "fake-token1")
method2, err := upsertTestAuthMethod(codec, "root", "dc1", testSessionID2)
require.NoError(t, err)
i2_r1, err := upsertTestBindingRule(
codec, "root", "dc1",
method2.Name,
"serviceaccount.name==abc",
structs.BindingRuleBindTypeService,
"abc",
)
require.NoError(t, err)
i2_r2, err := upsertTestBindingRule(
codec, "root", "dc1",
method2.Name,
"serviceaccount.name==def",
structs.BindingRuleBindTypeService,
"def",
)
require.NoError(t, err)
i2_t1 := createToken(method2.Name, "fake-token2")
i2_t2 := createToken(method2.Name, "fake-token2")
acl := ACL{srv: s1}
req := structs.ACLAuthMethodDeleteRequest{
Datacenter: "dc1",
AuthMethodName: method1.Name,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var ignored bool
err = acl.AuthMethodDelete(&req, &ignored)
require.NoError(t, err)
// Make sure the method is gone.
methodResp, err := retrieveTestAuthMethod(codec, "root", "dc1", method1.Name)
require.NoError(t, err)
require.Nil(t, methodResp.AuthMethod)
// Make sure the rules and tokens are gone.
for _, id := range []string{i1_r1.ID, i1_r2.ID} {
ruleResp, err := retrieveTestBindingRule(codec, "root", "dc1", id)
require.NoError(t, err)
require.Nil(t, ruleResp.BindingRule)
}
for _, id := range []string{i1_t1.AccessorID, i1_t2.AccessorID} {
tokResp, err := retrieveTestToken(codec, "root", "dc1", id)
require.NoError(t, err)
require.Nil(t, tokResp.Token)
}
// Make sure the rules and tokens for the untouched auth method are still there.
for _, id := range []string{i2_r1.ID, i2_r2.ID} {
ruleResp, err := retrieveTestBindingRule(codec, "root", "dc1", id)
require.NoError(t, err)
require.NotNil(t, ruleResp.BindingRule)
}
for _, id := range []string{i2_t1.AccessorID, i2_t2.AccessorID} {
tokResp, err := retrieveTestToken(codec, "root", "dc1", id)
require.NoError(t, err)
require.NotNil(t, tokResp.Token)
}
}
func TestACLEndpoint_AuthMethodList(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
i1, err := upsertTestAuthMethod(codec, "root", "dc1", "")
require.NoError(t, err)
i2, err := upsertTestAuthMethod(codec, "root", "dc1", "")
require.NoError(t, err)
acl := ACL{srv: s1}
req := structs.ACLAuthMethodListRequest{
Datacenter: "dc1",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLAuthMethodListResponse{}
err = acl.AuthMethodList(&req, &resp)
require.NoError(t, err)
require.ElementsMatch(t, gatherIDs(t, resp.AuthMethods), []string{i1.Name, i2.Name})
}
func TestACLEndpoint_BindingRuleSet(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
var ruleID string
testAuthMethod, err := upsertTestAuthMethod(codec, "root", "dc1", "")
require.NoError(t, err)
otherTestAuthMethod, err := upsertTestAuthMethod(codec, "root", "dc1", "")
require.NoError(t, err)
newRule := func() structs.ACLBindingRule {
return structs.ACLBindingRule{
Description: "foobar",
AuthMethod: testAuthMethod.Name,
Selector: "serviceaccount.name==abc",
BindType: structs.BindingRuleBindTypeService,
BindName: "abc",
}
}
requireSetErrors := func(t *testing.T, reqRule structs.ACLBindingRule) {
req := structs.ACLBindingRuleSetRequest{
Datacenter: "dc1",
BindingRule: reqRule,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLBindingRule{}
err := acl.BindingRuleSet(&req, &resp)
require.Error(t, err)
}
requireOK := func(t *testing.T, reqRule structs.ACLBindingRule) *structs.ACLBindingRule {
req := structs.ACLBindingRuleSetRequest{
Datacenter: "dc1",
BindingRule: reqRule,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLBindingRule{}
err := acl.BindingRuleSet(&req, &resp)
require.NoError(t, err)
require.NotEmpty(t, resp.ID)
return &resp
}
t.Run("Create it", func(t *testing.T) {
reqRule := newRule()
req := structs.ACLBindingRuleSetRequest{
Datacenter: "dc1",
BindingRule: reqRule,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLBindingRule{}
err := acl.BindingRuleSet(&req, &resp)
require.NoError(t, err)
require.NotNil(t, resp.ID)
// Get the rule directly to validate that it exists
ruleResp, err := retrieveTestBindingRule(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
rule := ruleResp.BindingRule
require.NotEmpty(t, rule.ID)
require.Equal(t, rule.Description, "foobar")
require.Equal(t, rule.AuthMethod, testAuthMethod.Name)
require.Equal(t, "serviceaccount.name==abc", rule.Selector)
require.Equal(t, structs.BindingRuleBindTypeService, rule.BindType)
require.Equal(t, "abc", rule.BindName)
ruleID = rule.ID
})
t.Run("Update fails; cannot change method name", func(t *testing.T) {
reqRule := newRule()
reqRule.ID = ruleID
reqRule.AuthMethod = otherTestAuthMethod.Name
requireSetErrors(t, reqRule)
})
t.Run("Update it - omit method name", func(t *testing.T) {
reqRule := newRule()
reqRule.ID = ruleID
reqRule.Description = "foobar modified 1"
reqRule.Selector = "serviceaccount.namespace==def"
reqRule.BindType = structs.BindingRuleBindTypeRole
reqRule.BindName = "def"
reqRule.AuthMethod = "" // clear
req := structs.ACLBindingRuleSetRequest{
Datacenter: "dc1",
BindingRule: reqRule,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLBindingRule{}
err := acl.BindingRuleSet(&req, &resp)
require.NoError(t, err)
require.NotNil(t, resp.ID)
// Get the rule directly to validate that it exists
ruleResp, err := retrieveTestBindingRule(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
rule := ruleResp.BindingRule
require.NotEmpty(t, rule.ID)
require.Equal(t, rule.Description, "foobar modified 1")
require.Equal(t, rule.AuthMethod, testAuthMethod.Name)
require.Equal(t, "serviceaccount.namespace==def", rule.Selector)
require.Equal(t, structs.BindingRuleBindTypeRole, rule.BindType)
require.Equal(t, "def", rule.BindName)
})
t.Run("Update it - specify method name", func(t *testing.T) {
reqRule := newRule()
reqRule.ID = ruleID
reqRule.Description = "foobar modified 2"
reqRule.Selector = "serviceaccount.namespace==def"
reqRule.BindType = structs.BindingRuleBindTypeRole
reqRule.BindName = "def"
req := structs.ACLBindingRuleSetRequest{
Datacenter: "dc1",
BindingRule: reqRule,
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLBindingRule{}
err := acl.BindingRuleSet(&req, &resp)
require.NoError(t, err)
require.NotNil(t, resp.ID)
// Get the rule directly to validate that it exists
ruleResp, err := retrieveTestBindingRule(codec, "root", "dc1", resp.ID)
require.NoError(t, err)
rule := ruleResp.BindingRule
require.NotEmpty(t, rule.ID)
require.Equal(t, rule.Description, "foobar modified 2")
require.Equal(t, rule.AuthMethod, testAuthMethod.Name)
require.Equal(t, "serviceaccount.namespace==def", rule.Selector)
require.Equal(t, structs.BindingRuleBindTypeRole, rule.BindType)
require.Equal(t, "def", rule.BindName)
})
t.Run("Create fails; empty method name", func(t *testing.T) {
reqRule := newRule()
reqRule.AuthMethod = ""
requireSetErrors(t, reqRule)
})
t.Run("Create fails; unknown method name", func(t *testing.T) {
reqRule := newRule()
reqRule.AuthMethod = "unknown"
requireSetErrors(t, reqRule)
})
t.Run("Create with no explicit selector", func(t *testing.T) {
reqRule := newRule()
reqRule.Selector = ""
rule := requireOK(t, reqRule)
require.Empty(t, rule.Selector, 0)
})
t.Run("Create fails; match selector with unknown vars", func(t *testing.T) {
reqRule := newRule()
reqRule.Selector = "serviceaccount.name==a and serviceaccount.bizarroname==b"
requireSetErrors(t, reqRule)
})
t.Run("Create fails; match selector invalid", func(t *testing.T) {
reqRule := newRule()
reqRule.Selector = "serviceaccount.name"
requireSetErrors(t, reqRule)
})
t.Run("Create fails; empty bind type", func(t *testing.T) {
reqRule := newRule()
reqRule.BindType = ""
requireSetErrors(t, reqRule)
})
t.Run("Create fails; empty bind name", func(t *testing.T) {
reqRule := newRule()
reqRule.BindName = ""
requireSetErrors(t, reqRule)
})
t.Run("Create fails; invalid bind type", func(t *testing.T) {
reqRule := newRule()
reqRule.BindType = "invalid"
requireSetErrors(t, reqRule)
})
t.Run("Create fails; bind name with unknown vars", func(t *testing.T) {
reqRule := newRule()
reqRule.BindName = "method-${serviceaccount.bizarroname}"
requireSetErrors(t, reqRule)
})
t.Run("Create fails; invalid bind name no template", func(t *testing.T) {
reqRule := newRule()
reqRule.BindName = "-abc:"
requireSetErrors(t, reqRule)
})
t.Run("Create fails; invalid bind name with template", func(t *testing.T) {
reqRule := newRule()
reqRule.BindName = "method-${serviceaccount.name"
requireSetErrors(t, reqRule)
})
t.Run("Create fails; invalid bind name after template computed", func(t *testing.T) {
reqRule := newRule()
reqRule.BindName = "method-${serviceaccount.name}:blah-"
requireSetErrors(t, reqRule)
})
}
func TestACLEndpoint_BindingRuleDelete(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
testAuthMethod, err := upsertTestAuthMethod(codec, "root", "dc1", "")
require.NoError(t, err)
existingRule, err := upsertTestBindingRule(
codec, "root", "dc1",
testAuthMethod.Name,
"serviceaccount.name==abc",
structs.BindingRuleBindTypeService,
"abc",
)
require.NoError(t, err)
acl := ACL{srv: s1}
t.Run("normal", func(t *testing.T) {
req := structs.ACLBindingRuleDeleteRequest{
Datacenter: "dc1",
BindingRuleID: existingRule.ID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var ignored bool
err = acl.BindingRuleDelete(&req, &ignored)
require.NoError(t, err)
// Make sure the rule is gone
ruleResp, err := retrieveTestBindingRule(codec, "root", "dc1", existingRule.ID)
require.NoError(t, err)
require.Nil(t, ruleResp.BindingRule)
})
t.Run("delete something that doesn't exist", func(t *testing.T) {
fakeID, err := uuid.GenerateUUID()
require.NoError(t, err)
req := structs.ACLBindingRuleDeleteRequest{
Datacenter: "dc1",
BindingRuleID: fakeID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var ignored bool
err = acl.BindingRuleDelete(&req, &ignored)
require.NoError(t, err)
})
}
func TestACLEndpoint_BindingRuleList(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
testAuthMethod, err := upsertTestAuthMethod(codec, "root", "dc1", "")
require.NoError(t, err)
r1, err := upsertTestBindingRule(
codec, "root", "dc1",
testAuthMethod.Name,
"serviceaccount.name==abc",
structs.BindingRuleBindTypeService,
"abc",
)
require.NoError(t, err)
r2, err := upsertTestBindingRule(
codec, "root", "dc1",
testAuthMethod.Name,
"serviceaccount.name==def",
structs.BindingRuleBindTypeService,
"def",
)
require.NoError(t, err)
acl := ACL{srv: s1}
req := structs.ACLBindingRuleListRequest{
Datacenter: "dc1",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLBindingRuleListResponse{}
err = acl.BindingRuleList(&req, &resp)
require.NoError(t, err)
require.ElementsMatch(t, gatherIDs(t, resp.BindingRules), []string{r1.ID, r2.ID})
}
func TestACLEndpoint_SecureIntroEndpoints_LocalTokensDisabled(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
dir2, s2 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.Datacenter = "dc2"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
// disable local tokens
c.ACLTokenReplication = false
})
defer os.RemoveAll(dir2)
defer s2.Shutdown()
codec2 := rpcClient(t, s2)
defer codec2.Close()
s2.tokens.UpdateReplicationToken("root", tokenStore.TokenSourceConfig)
testrpc.WaitForLeader(t, s1.RPC, "dc1")
testrpc.WaitForLeader(t, s2.RPC, "dc2")
// Try to join
joinWAN(t, s2, s1)
waitForNewACLs(t, s1)
waitForNewACLs(t, s2)
acl2 := ACL{srv: s2}
var ignored bool
errString := errAuthMethodsRequireTokenReplication.Error()
t.Run("AuthMethodRead", func(t *testing.T) {
requireErrorContains(t,
acl2.AuthMethodRead(&structs.ACLAuthMethodGetRequest{Datacenter: "dc2"},
&structs.ACLAuthMethodResponse{}),
errString,
)
})
t.Run("AuthMethodSet", func(t *testing.T) {
requireErrorContains(t,
acl2.AuthMethodSet(&structs.ACLAuthMethodSetRequest{Datacenter: "dc2"},
&structs.ACLAuthMethod{}),
errString,
)
})
t.Run("AuthMethodDelete", func(t *testing.T) {
requireErrorContains(t,
acl2.AuthMethodDelete(&structs.ACLAuthMethodDeleteRequest{Datacenter: "dc2"}, &ignored),
errString,
)
})
t.Run("AuthMethodList", func(t *testing.T) {
requireErrorContains(t,
acl2.AuthMethodList(&structs.ACLAuthMethodListRequest{Datacenter: "dc2"},
&structs.ACLAuthMethodListResponse{}),
errString,
)
})
t.Run("BindingRuleRead", func(t *testing.T) {
requireErrorContains(t,
acl2.BindingRuleRead(&structs.ACLBindingRuleGetRequest{Datacenter: "dc2"},
&structs.ACLBindingRuleResponse{}),
errString,
)
})
t.Run("BindingRuleSet", func(t *testing.T) {
requireErrorContains(t,
acl2.BindingRuleSet(&structs.ACLBindingRuleSetRequest{Datacenter: "dc2"},
&structs.ACLBindingRule{}),
errString,
)
})
t.Run("BindingRuleDelete", func(t *testing.T) {
requireErrorContains(t,
acl2.BindingRuleDelete(&structs.ACLBindingRuleDeleteRequest{Datacenter: "dc2"}, &ignored),
errString,
)
})
t.Run("BindingRuleList", func(t *testing.T) {
requireErrorContains(t,
acl2.BindingRuleList(&structs.ACLBindingRuleListRequest{Datacenter: "dc2"},
&structs.ACLBindingRuleListResponse{}),
errString,
)
})
t.Run("Login", func(t *testing.T) {
requireErrorContains(t,
acl2.Login(&structs.ACLLoginRequest{Datacenter: "dc2"},
&structs.ACLToken{}),
errString,
)
})
t.Run("Logout", func(t *testing.T) {
requireErrorContains(t,
acl2.Logout(&structs.ACLLogoutRequest{Datacenter: "dc2"}, &ignored),
errString,
)
})
}
func TestACLEndpoint_SecureIntroEndpoints_OnlyCreateLocalData(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec1 := rpcClient(t, s1)
defer codec1.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
dir2, s2 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.Datacenter = "dc2"
c.ACLTokenMinExpirationTTL = 10 * time.Millisecond
c.ACLTokenMaxExpirationTTL = 5 * time.Second
// enable token replication so secure intro works
c.ACLTokenReplication = true
})
defer os.RemoveAll(dir2)
defer s2.Shutdown()
codec2 := rpcClient(t, s2)
defer codec2.Close()
s2.tokens.UpdateReplicationToken("root", tokenStore.TokenSourceConfig)
testrpc.WaitForLeader(t, s1.RPC, "dc1")
testrpc.WaitForLeader(t, s2.RPC, "dc2")
// Try to join
joinWAN(t, s2, s1)
waitForNewACLs(t, s1)
waitForNewACLs(t, s2)
// Ensure s2 is authoritative.
waitForNewACLReplication(t, s2, structs.ACLReplicateTokens, 1, 1, 0)
acl := ACL{srv: s1}
acl2 := ACL{srv: s2}
//
// this order is specific so that we can do it in one pass
//
testSessionID_1 := testauth.StartSession()
defer testauth.ResetSession(testSessionID_1)
testSessionID_2 := testauth.StartSession()
defer testauth.ResetSession(testSessionID_2)
testauth.InstallSessionToken(
testSessionID_1,
"fake-web1-token",
"default", "web1", "abc123",
)
testauth.InstallSessionToken(
testSessionID_2,
"fake-web2-token",
"default", "web2", "def456",
)
t.Run("create auth method", func(t *testing.T) {
req := structs.ACLAuthMethodSetRequest{
Datacenter: "dc2",
AuthMethod: structs.ACLAuthMethod{
Name: "testmethod",
Description: "test original",
Type: "testing",
Config: map[string]interface{}{
"SessionID": testSessionID_2,
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLAuthMethod{}
require.NoError(t, acl2.AuthMethodSet(&req, &resp))
// present in dc2
resp2, err := retrieveTestAuthMethod(codec2, "root", "dc2", "testmethod")
require.NoError(t, err)
require.NotNil(t, resp2.AuthMethod)
require.Equal(t, "test original", resp2.AuthMethod.Description)
// absent in dc1
resp2, err = retrieveTestAuthMethod(codec1, "root", "dc1", "testmethod")
require.NoError(t, err)
require.Nil(t, resp2.AuthMethod)
})
t.Run("update auth method", func(t *testing.T) {
req := structs.ACLAuthMethodSetRequest{
Datacenter: "dc2",
AuthMethod: structs.ACLAuthMethod{
Name: "testmethod",
Description: "test updated",
Config: map[string]interface{}{
"SessionID": testSessionID_2,
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLAuthMethod{}
require.NoError(t, acl2.AuthMethodSet(&req, &resp))
// present in dc2
resp2, err := retrieveTestAuthMethod(codec2, "root", "dc2", "testmethod")
require.NoError(t, err)
require.NotNil(t, resp2.AuthMethod)
require.Equal(t, "test updated", resp2.AuthMethod.Description)
// absent in dc1
resp2, err = retrieveTestAuthMethod(codec1, "root", "dc1", "testmethod")
require.NoError(t, err)
require.Nil(t, resp2.AuthMethod)
})
t.Run("read auth method", func(t *testing.T) {
// present in dc2
req := structs.ACLAuthMethodGetRequest{
Datacenter: "dc2",
AuthMethodName: "testmethod",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLAuthMethodResponse{}
require.NoError(t, acl2.AuthMethodRead(&req, &resp))
require.NotNil(t, resp.AuthMethod)
require.Equal(t, "test updated", resp.AuthMethod.Description)
// absent in dc1
req = structs.ACLAuthMethodGetRequest{
Datacenter: "dc1",
AuthMethodName: "testmethod",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp = structs.ACLAuthMethodResponse{}
require.NoError(t, acl.AuthMethodRead(&req, &resp))
require.Nil(t, resp.AuthMethod)
})
t.Run("list auth method", func(t *testing.T) {
// present in dc2
req := structs.ACLAuthMethodListRequest{
Datacenter: "dc2",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLAuthMethodListResponse{}
require.NoError(t, acl2.AuthMethodList(&req, &resp))
require.Len(t, resp.AuthMethods, 1)
// absent in dc1
req = structs.ACLAuthMethodListRequest{
Datacenter: "dc1",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp = structs.ACLAuthMethodListResponse{}
require.NoError(t, acl.AuthMethodList(&req, &resp))
require.Len(t, resp.AuthMethods, 0)
})
var ruleID string
t.Run("create binding rule", func(t *testing.T) {
req := structs.ACLBindingRuleSetRequest{
Datacenter: "dc2",
BindingRule: structs.ACLBindingRule{
Description: "test original",
AuthMethod: "testmethod",
BindType: structs.BindingRuleBindTypeService,
BindName: "${serviceaccount.name}",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLBindingRule{}
require.NoError(t, acl2.BindingRuleSet(&req, &resp))
ruleID = resp.ID
// present in dc2
resp2, err := retrieveTestBindingRule(codec2, "root", "dc2", ruleID)
require.NoError(t, err)
require.NotNil(t, resp2.BindingRule)
require.Equal(t, "test original", resp2.BindingRule.Description)
// absent in dc1
resp2, err = retrieveTestBindingRule(codec1, "root", "dc1", ruleID)
require.NoError(t, err)
require.Nil(t, resp2.BindingRule)
})
t.Run("update binding rule", func(t *testing.T) {
req := structs.ACLBindingRuleSetRequest{
Datacenter: "dc2",
BindingRule: structs.ACLBindingRule{
ID: ruleID,
Description: "test updated",
AuthMethod: "testmethod",
BindType: structs.BindingRuleBindTypeService,
BindName: "${serviceaccount.name}",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
resp := structs.ACLBindingRule{}
require.NoError(t, acl2.BindingRuleSet(&req, &resp))
ruleID = resp.ID
// present in dc2
resp2, err := retrieveTestBindingRule(codec2, "root", "dc2", ruleID)
require.NoError(t, err)
require.NotNil(t, resp2.BindingRule)
require.Equal(t, "test updated", resp2.BindingRule.Description)
// absent in dc1
resp2, err = retrieveTestBindingRule(codec1, "root", "dc1", ruleID)
require.NoError(t, err)
require.Nil(t, resp2.BindingRule)
})
t.Run("read binding rule", func(t *testing.T) {
// present in dc2
req := structs.ACLBindingRuleGetRequest{
Datacenter: "dc2",
BindingRuleID: ruleID,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLBindingRuleResponse{}
require.NoError(t, acl2.BindingRuleRead(&req, &resp))
require.NotNil(t, resp.BindingRule)
require.Equal(t, "test updated", resp.BindingRule.Description)
// absent in dc1
req = structs.ACLBindingRuleGetRequest{
Datacenter: "dc1",
BindingRuleID: ruleID,
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp = structs.ACLBindingRuleResponse{}
require.NoError(t, acl.BindingRuleRead(&req, &resp))
require.Nil(t, resp.BindingRule)
})
t.Run("list binding rule", func(t *testing.T) {
// present in dc2
req := structs.ACLBindingRuleListRequest{
Datacenter: "dc2",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.ACLBindingRuleListResponse{}
require.NoError(t, acl2.BindingRuleList(&req, &resp))
require.Len(t, resp.BindingRules, 1)
// absent in dc1
req = structs.ACLBindingRuleListRequest{
Datacenter: "dc1",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp = structs.ACLBindingRuleListResponse{}
require.NoError(t, acl.BindingRuleList(&req, &resp))
require.Len(t, resp.BindingRules, 0)
})
var remoteToken *structs.ACLToken
t.Run("login in remote", func(t *testing.T) {
req := structs.ACLLoginRequest{
Datacenter: "dc2",
Auth: &structs.ACLLoginParams{
AuthMethod: "testmethod",
BearerToken: "fake-web2-token",
},
}
resp := structs.ACLToken{}
require.NoError(t, acl2.Login(&req, &resp))
remoteToken = &resp
// present in dc2
resp2, err := retrieveTestToken(codec2, "root", "dc2", remoteToken.AccessorID)
require.NoError(t, err)
require.NotNil(t, resp2.Token)
require.Len(t, resp2.Token.ServiceIdentities, 1)
require.Equal(t, "web2", resp2.Token.ServiceIdentities[0].ServiceName)
// absent in dc1
resp2, err = retrieveTestToken(codec1, "root", "dc1", remoteToken.AccessorID)
require.NoError(t, err)
require.Nil(t, resp2.Token)
})
// We delay until now to setup an auth method and binding rule in the
// primary so our earlier listing tests were sane. We need to be able to
// use auth methods in both datacenters in order to verify Logout is
// properly scoped.
t.Run("initialize primary so we can test logout", func(t *testing.T) {
reqAM := structs.ACLAuthMethodSetRequest{
Datacenter: "dc1",
AuthMethod: structs.ACLAuthMethod{
Name: "primarymethod",
Type: "testing",
Config: map[string]interface{}{
"SessionID": testSessionID_1,
},
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
respAM := structs.ACLAuthMethod{}
require.NoError(t, acl.AuthMethodSet(&reqAM, &respAM))
reqBR := structs.ACLBindingRuleSetRequest{
Datacenter: "dc1",
BindingRule: structs.ACLBindingRule{
AuthMethod: "primarymethod",
BindType: structs.BindingRuleBindTypeService,
BindName: "${serviceaccount.name}",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
respBR := structs.ACLBindingRule{}
require.NoError(t, acl.BindingRuleSet(&reqBR, &respBR))
})
var primaryToken *structs.ACLToken
t.Run("login in primary", func(t *testing.T) {
req := structs.ACLLoginRequest{
Datacenter: "dc1",
Auth: &structs.ACLLoginParams{
AuthMethod: "primarymethod",
BearerToken: "fake-web1-token",
},
}
resp := structs.ACLToken{}
require.NoError(t, acl.Login(&req, &resp))
primaryToken = &resp
// present in dc1
resp2, err := retrieveTestToken(codec1, "root", "dc1", primaryToken.AccessorID)
require.NoError(t, err)
require.NotNil(t, resp2.Token)
require.Len(t, resp2.Token.ServiceIdentities, 1)
require.Equal(t, "web1", resp2.Token.ServiceIdentities[0].ServiceName)
// absent in dc2
resp2, err = retrieveTestToken(codec2, "root", "dc2", primaryToken.AccessorID)
require.NoError(t, err)
require.Nil(t, resp2.Token)
})
t.Run("logout of remote token in remote dc", func(t *testing.T) {
2019-11-01 20:11:44 +00:00
// if the other test fails this one will to but will now not segfault
require.NotNil(t, remoteToken)
req := structs.ACLLogoutRequest{
Datacenter: "dc2",
WriteRequest: structs.WriteRequest{Token: remoteToken.SecretID},
}
var ignored bool
require.NoError(t, acl.Logout(&req, &ignored))
// absent in dc2
resp2, err := retrieveTestToken(codec2, "root", "dc2", remoteToken.AccessorID)
require.NoError(t, err)
require.Nil(t, resp2.Token)
// absent in dc1
resp2, err = retrieveTestToken(codec1, "root", "dc1", remoteToken.AccessorID)
require.NoError(t, err)
require.Nil(t, resp2.Token)
})
t.Run("logout of primary token in remote dc should not work", func(t *testing.T) {
req := structs.ACLLogoutRequest{
Datacenter: "dc2",
WriteRequest: structs.WriteRequest{Token: primaryToken.SecretID},
}
var ignored bool
requireErrorContains(t, acl.Logout(&req, &ignored), "ACL not found")
// present in dc1
resp2, err := retrieveTestToken(codec1, "root", "dc1", primaryToken.AccessorID)
require.NoError(t, err)
require.NotNil(t, resp2.Token)
require.Len(t, resp2.Token.ServiceIdentities, 1)
require.Equal(t, "web1", resp2.Token.ServiceIdentities[0].ServiceName)
// absent in dc2
resp2, err = retrieveTestToken(codec2, "root", "dc2", primaryToken.AccessorID)
require.NoError(t, err)
require.Nil(t, resp2.Token)
})
// Don't trigger the auth method delete cascade so we know the individual
// endpoints follow the rules.
t.Run("delete binding rule", func(t *testing.T) {
req := structs.ACLBindingRuleDeleteRequest{
Datacenter: "dc2",
BindingRuleID: ruleID,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var ignored bool
require.NoError(t, acl2.BindingRuleDelete(&req, &ignored))
// absent in dc2
resp2, err := retrieveTestBindingRule(codec2, "root", "dc2", ruleID)
require.NoError(t, err)
require.Nil(t, resp2.BindingRule)
// absent in dc1
resp2, err = retrieveTestBindingRule(codec1, "root", "dc1", ruleID)
require.NoError(t, err)
require.Nil(t, resp2.BindingRule)
})
t.Run("delete auth method", func(t *testing.T) {
req := structs.ACLAuthMethodDeleteRequest{
Datacenter: "dc2",
AuthMethodName: "testmethod",
WriteRequest: structs.WriteRequest{Token: "root"},
}
var ignored bool
require.NoError(t, acl2.AuthMethodDelete(&req, &ignored))
// absent in dc2
resp2, err := retrieveTestAuthMethod(codec2, "root", "dc2", "testmethod")
require.NoError(t, err)
require.Nil(t, resp2.AuthMethod)
// absent in dc1
resp2, err = retrieveTestAuthMethod(codec1, "root", "dc1", "testmethod")
require.NoError(t, err)
require.Nil(t, resp2.AuthMethod)
})
}
func TestACLEndpoint_Login(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
testSessionID := testauth.StartSession()
defer testauth.ResetSession(testSessionID)
testauth.InstallSessionToken(
testSessionID,
"fake-web", // no rules
"default", "web", "abc123",
)
testauth.InstallSessionToken(
testSessionID,
"fake-db", // 1 rule (service)
"default", "db", "def456",
)
testauth.InstallSessionToken(
testSessionID,
"fake-vault", // 1 rule (role)
"default", "vault", "jkl012",
)
testauth.InstallSessionToken(
testSessionID,
"fake-monolith", // 2 rules (one of each)
"default", "monolith", "ghi789",
)
method, err := upsertTestAuthMethod(codec, "root", "dc1", testSessionID)
require.NoError(t, err)
// 'fake-db' rules
ruleDB, err := upsertTestBindingRule(
codec, "root", "dc1", method.Name,
"serviceaccount.namespace==default and serviceaccount.name==db",
structs.BindingRuleBindTypeService,
"method-${serviceaccount.name}",
)
require.NoError(t, err)
// 'fake-vault' rules
_, err = upsertTestBindingRule(
codec, "root", "dc1", method.Name,
"serviceaccount.namespace==default and serviceaccount.name==vault",
structs.BindingRuleBindTypeRole,
"method-${serviceaccount.name}",
)
require.NoError(t, err)
// 'fake-monolith' rules
_, err = upsertTestBindingRule(
codec, "root", "dc1", method.Name,
"serviceaccount.namespace==default and serviceaccount.name==monolith",
structs.BindingRuleBindTypeService,
"method-${serviceaccount.name}",
)
require.NoError(t, err)
_, err = upsertTestBindingRule(
codec, "root", "dc1", method.Name,
"serviceaccount.namespace==default and serviceaccount.name==monolith",
structs.BindingRuleBindTypeRole,
"method-${serviceaccount.name}",
)
require.NoError(t, err)
t.Run("do not provide a token", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-web",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
req.Token = "nope"
resp := structs.ACLToken{}
requireErrorContains(t, acl.Login(&req, &resp), "do not provide a token")
})
t.Run("unknown method", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name + "-notexist",
BearerToken: "fake-web",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
requireErrorContains(t, acl.Login(&req, &resp), "ACL not found")
})
t.Run("invalid method token", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "invalid",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
require.Error(t, acl.Login(&req, &resp))
})
t.Run("valid method token no bindings", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-web",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
requireErrorContains(t, acl.Login(&req, &resp), "Permission denied")
})
t.Run("valid method token 1 role binding and role does not exist", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-vault",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
requireErrorContains(t, acl.Login(&req, &resp), "Permission denied")
})
// create the role so that the bindtype=role login works
var vaultRoleID string
{
arg := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
Name: "method-vault",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
var out structs.ACLRole
require.NoError(t, acl.RoleSet(&arg, &out))
vaultRoleID = out.ID
}
t.Run("valid method token 1 role binding and role now exists", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-vault",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
require.NoError(t, acl.Login(&req, &resp))
require.Equal(t, method.Name, resp.AuthMethod)
require.Equal(t, `token created via login: {"pod":"pod1"}`, resp.Description)
require.True(t, resp.Local)
require.Len(t, resp.ServiceIdentities, 0)
require.Len(t, resp.Roles, 1)
role := resp.Roles[0]
require.Equal(t, vaultRoleID, role.ID)
require.Equal(t, "method-vault", role.Name)
})
t.Run("valid method token 1 service binding 1 role binding and role does not exist", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-monolith",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
require.NoError(t, acl.Login(&req, &resp))
require.Equal(t, method.Name, resp.AuthMethod)
require.Equal(t, `token created via login: {"pod":"pod1"}`, resp.Description)
require.True(t, resp.Local)
require.Len(t, resp.ServiceIdentities, 1)
require.Len(t, resp.Roles, 0)
svcid := resp.ServiceIdentities[0]
require.Len(t, svcid.Datacenters, 0)
require.Equal(t, "method-monolith", svcid.ServiceName)
})
// create the role so that the bindtype=role login works
var monolithRoleID string
{
arg := structs.ACLRoleSetRequest{
Datacenter: "dc1",
Role: structs.ACLRole{
Name: "method-monolith",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
var out structs.ACLRole
require.NoError(t, acl.RoleSet(&arg, &out))
monolithRoleID = out.ID
}
t.Run("valid method token 1 service binding 1 role binding and role now exists", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-monolith",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
require.NoError(t, acl.Login(&req, &resp))
require.Equal(t, method.Name, resp.AuthMethod)
require.Equal(t, `token created via login: {"pod":"pod1"}`, resp.Description)
require.True(t, resp.Local)
require.Len(t, resp.ServiceIdentities, 1)
require.Len(t, resp.Roles, 1)
role := resp.Roles[0]
require.Equal(t, monolithRoleID, role.ID)
require.Equal(t, "method-monolith", role.Name)
svcid := resp.ServiceIdentities[0]
require.Len(t, svcid.Datacenters, 0)
require.Equal(t, "method-monolith", svcid.ServiceName)
})
t.Run("valid bearer token 1 service binding", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-db",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
require.NoError(t, acl.Login(&req, &resp))
require.Equal(t, method.Name, resp.AuthMethod)
require.Equal(t, `token created via login: {"pod":"pod1"}`, resp.Description)
require.True(t, resp.Local)
require.Len(t, resp.Roles, 0)
require.Len(t, resp.ServiceIdentities, 1)
svcid := resp.ServiceIdentities[0]
require.Len(t, svcid.Datacenters, 0)
require.Equal(t, "method-db", svcid.ServiceName)
})
{
req := structs.ACLBindingRuleSetRequest{
Datacenter: "dc1",
BindingRule: structs.ACLBindingRule{
AuthMethod: ruleDB.AuthMethod,
BindType: structs.BindingRuleBindTypeService,
BindName: ruleDB.BindName,
Selector: "",
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
var out structs.ACLBindingRule
require.NoError(t, acl.BindingRuleSet(&req, &out))
}
t.Run("valid bearer token 1 binding (no selectors this time)", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-db",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
require.NoError(t, acl.Login(&req, &resp))
require.Equal(t, method.Name, resp.AuthMethod)
require.Equal(t, `token created via login: {"pod":"pod1"}`, resp.Description)
require.True(t, resp.Local)
require.Len(t, resp.Roles, 0)
require.Len(t, resp.ServiceIdentities, 1)
svcid := resp.ServiceIdentities[0]
require.Len(t, svcid.Datacenters, 0)
require.Equal(t, "method-db", svcid.ServiceName)
})
testSessionID_2 := testauth.StartSession()
defer testauth.ResetSession(testSessionID_2)
{
// Update the method to force the cache to invalidate for the next
// subtest.
updated := *method
updated.Description = "updated for the test"
updated.Config = map[string]interface{}{
"SessionID": testSessionID_2,
}
req := structs.ACLAuthMethodSetRequest{
Datacenter: "dc1",
AuthMethod: updated,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var ignored structs.ACLAuthMethod
require.NoError(t, acl.AuthMethodSet(&req, &ignored))
}
t.Run("updating the method invalidates the cache", func(t *testing.T) {
// We'll try to login with the 'fake-db' cred which DOES exist in the
// old fake validator, but no longer exists in the new fake validator.
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-db",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
requireErrorContains(t, acl.Login(&req, &resp), "ACL not found")
})
}
func TestACLEndpoint_Login_k8s(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
// spin up a fake api server
testSrv := kubeauth.StartTestAPIServer(t)
defer testSrv.Stop()
testSrv.AuthorizeJWT(goodJWT_A)
testSrv.SetAllowedServiceAccount(
"default",
"demo",
"76091af4-4b56-11e9-ac4b-708b11801cbe",
"",
goodJWT_B,
)
method, err := upsertTestKubernetesAuthMethod(
codec, "root", "dc1",
testSrv.CACert(),
testSrv.Addr(),
goodJWT_A,
)
require.NoError(t, err)
t.Run("invalid bearer token", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "invalid",
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
require.Error(t, acl.Login(&req, &resp))
})
t.Run("valid bearer token no bindings", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: goodJWT_B,
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
requireErrorContains(t, acl.Login(&req, &resp), "Permission denied")
})
_, err = upsertTestBindingRule(
codec, "root", "dc1", method.Name,
"serviceaccount.namespace==default",
structs.BindingRuleBindTypeService,
"${serviceaccount.name}",
)
require.NoError(t, err)
t.Run("valid bearer token 1 service binding", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: goodJWT_B,
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
require.NoError(t, acl.Login(&req, &resp))
require.Equal(t, method.Name, resp.AuthMethod)
require.Equal(t, `token created via login: {"pod":"pod1"}`, resp.Description)
require.True(t, resp.Local)
require.Len(t, resp.Roles, 0)
require.Len(t, resp.ServiceIdentities, 1)
svcid := resp.ServiceIdentities[0]
require.Len(t, svcid.Datacenters, 0)
require.Equal(t, "demo", svcid.ServiceName)
})
// annotate the account
testSrv.SetAllowedServiceAccount(
"default",
"demo",
"76091af4-4b56-11e9-ac4b-708b11801cbe",
"alternate-name",
goodJWT_B,
)
t.Run("valid bearer token 1 service binding - with annotation", func(t *testing.T) {
req := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: goodJWT_B,
Meta: map[string]string{"pod": "pod1"},
},
Datacenter: "dc1",
}
resp := structs.ACLToken{}
require.NoError(t, acl.Login(&req, &resp))
require.Equal(t, method.Name, resp.AuthMethod)
require.Equal(t, `token created via login: {"pod":"pod1"}`, resp.Description)
require.True(t, resp.Local)
require.Len(t, resp.Roles, 0)
require.Len(t, resp.ServiceIdentities, 1)
svcid := resp.ServiceIdentities[0]
require.Len(t, svcid.Datacenters, 0)
require.Equal(t, "alternate-name", svcid.ServiceName)
})
}
func TestACLEndpoint_Logout(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.ACLDatacenter = "dc1"
c.ACLsEnabled = true
c.ACLMasterToken = "root"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
acl := ACL{srv: s1}
testSessionID := testauth.StartSession()
defer testauth.ResetSession(testSessionID)
testauth.InstallSessionToken(
testSessionID,
"fake-db", // 1 rule
"default", "db", "def456",
)
method, err := upsertTestAuthMethod(codec, "root", "dc1", testSessionID)
require.NoError(t, err)
_, err = upsertTestBindingRule(
codec, "root", "dc1", method.Name,
"",
structs.BindingRuleBindTypeService,
"method-${serviceaccount.name}",
)
require.NoError(t, err)
t.Run("you must provide a token", func(t *testing.T) {
req := structs.ACLLogoutRequest{
Datacenter: "dc1",
// WriteRequest: structs.WriteRequest{Token: "root"},
}
req.Token = ""
var ignored bool
requireErrorContains(t, acl.Logout(&req, &ignored), "ACL not found")
})
t.Run("logout from deleted token", func(t *testing.T) {
req := structs.ACLLogoutRequest{
Datacenter: "dc1",
WriteRequest: structs.WriteRequest{Token: "not-found"},
}
var ignored bool
requireErrorContains(t, acl.Logout(&req, &ignored), "ACL not found")
})
t.Run("logout from non-auth method-linked token should fail", func(t *testing.T) {
req := structs.ACLLogoutRequest{
Datacenter: "dc1",
WriteRequest: structs.WriteRequest{Token: "root"},
}
var ignored bool
requireErrorContains(t, acl.Logout(&req, &ignored), "Permission denied")
})
t.Run("login then logout", func(t *testing.T) {
// Create a totally legit Login token.
loginReq := structs.ACLLoginRequest{
Auth: &structs.ACLLoginParams{
AuthMethod: method.Name,
BearerToken: "fake-db",
},
Datacenter: "dc1",
}
loginToken := structs.ACLToken{}
require.NoError(t, acl.Login(&loginReq, &loginToken))
require.NotEmpty(t, loginToken.SecretID)
// Now turn around and nuke it.
req := structs.ACLLogoutRequest{
Datacenter: "dc1",
WriteRequest: structs.WriteRequest{Token: loginToken.SecretID},
}
var ignored bool
require.NoError(t, acl.Logout(&req, &ignored))
})
}
func gatherIDs(t *testing.T, v interface{}) []string {
t.Helper()
var out []string
switch x := v.(type) {
case []*structs.ACLRole:
for _, r := range x {
out = append(out, r.ID)
}
case structs.ACLRoles:
for _, r := range x {
out = append(out, r.ID)
}
case []*structs.ACLPolicy:
for _, p := range x {
out = append(out, p.ID)
}
case structs.ACLPolicyListStubs:
for _, p := range x {
out = append(out, p.ID)
}
case []*structs.ACLToken:
for _, p := range x {
out = append(out, p.AccessorID)
}
case structs.ACLTokenListStubs:
for _, p := range x {
out = append(out, p.AccessorID)
}
case []*structs.ACLAuthMethod:
for _, p := range x {
out = append(out, p.Name)
}
case structs.ACLAuthMethodListStubs:
for _, p := range x {
out = append(out, p.Name)
}
case []*structs.ACLBindingRule:
for _, p := range x {
out = append(out, p.ID)
}
case structs.ACLBindingRules:
for _, p := range x {
out = append(out, p.ID)
}
default:
t.Fatalf("unknown type: %T", x)
}
return out
}
func TestValidateBindingRuleBindName(t *testing.T) {
t.Parallel()
type testcase struct {
name string
bindType string
bindName string
fields string
valid bool // valid HIL, invalid contents
err bool // invalid HIL
}
for _, test := range []testcase{
{"no bind type",
"", "", "", false, false},
{"bad bind type",
"invalid", "blah", "", false, true},
// valid HIL, invalid name
{"empty",
"both", "", "", false, false},
{"just end",
"both", "}", "", false, false},
{"var without start",
"both", " item }", "item", false, false},
{"two vars missing second start",
"both", "before-${ item }after--more }", "item,more", false, false},
// names for the two types are validated differently
{"@ is disallowed",
"both", "bad@name", "", false, false},
{"leading dash",
"role", "-name", "", true, false},
{"leading dash",
"service", "-name", "", false, false},
{"trailing dash",
"role", "name-", "", true, false},
{"trailing dash",
"service", "name-", "", false, false},
{"inner dash",
"both", "name-end", "", true, false},
{"upper case",
"role", "NAME", "", true, false},
{"upper case",
"service", "NAME", "", false, false},
// valid HIL, valid name
{"no vars",
"both", "nothing", "", true, false},
{"just var",
"both", "${item}", "item", true, false},
{"var in middle",
"both", "before-${item}after", "item", true, false},
{"two vars",
"both", "before-${item}after-${more}", "item,more", true, false},
// bad
{"no bind name",
"both", "", "", false, false},
{"just start",
"both", "${", "", false, true},
{"backwards",
"both", "}${", "", false, true},
{"no varname",
"both", "${}", "", false, true},
{"missing map key",
"both", "${item}", "", false, true},
{"var without end",
"both", "${ item ", "item", false, true},
{"two vars missing first end",
"both", "before-${ item after-${ more }", "item,more", false, true},
} {
var cases []testcase
if test.bindType == "both" {
test1 := test
test1.bindType = "role"
test2 := test
test2.bindType = "service"
cases = []testcase{test1, test2}
} else {
cases = []testcase{test}
}
for _, test := range cases {
test := test
t.Run(test.bindType+"--"+test.name, func(t *testing.T) {
t.Parallel()
valid, err := validateBindingRuleBindName(
test.bindType,
test.bindName,
strings.Split(test.fields, ","),
)
if test.err {
require.NotNil(t, err)
require.False(t, valid)
} else {
require.NoError(t, err)
require.Equal(t, test.valid, valid)
}
})
}
}
}
// upsertTestToken creates a token for testing purposes
func upsertTestToken(codec rpc.ClientCodec, masterToken string, datacenter string,
tokenModificationFn func(token *structs.ACLToken)) (*structs.ACLToken, error) {
arg := structs.ACLTokenSetRequest{
Datacenter: datacenter,
ACLToken: structs.ACLToken{
Description: "User token",
Local: false,
Policies: nil,
},
WriteRequest: structs.WriteRequest{Token: masterToken},
}
if tokenModificationFn != nil {
tokenModificationFn(&arg.ACLToken)
}
var out structs.ACLToken
err := msgpackrpc.CallWithCodec(codec, "ACL.TokenSet", &arg, &out)
if err != nil {
return nil, err
}
if out.AccessorID == "" {
return nil, fmt.Errorf("AccessorID is nil: %v", out)
}
return &out, nil
}
func upsertTestTokenWithPolicyRules(codec rpc.ClientCodec, masterToken string, datacenter string, rules string) (*structs.ACLToken, error) {
policy, err := upsertTestPolicyWithRules(codec, masterToken, datacenter, rules)
if err != nil {
return nil, err
}
token, err := upsertTestToken(codec, masterToken, datacenter, func(token *structs.ACLToken) {
token.Policies = []structs.ACLTokenPolicyLink{{ID: policy.ID}}
})
if err != nil {
return nil, err
}
return token, nil
}
func retrieveTestTokenAccessorForSecret(codec rpc.ClientCodec, masterToken string, datacenter string, id string) (string, error) {
arg := structs.ACLTokenGetRequest{
TokenID: id,
TokenIDType: structs.ACLTokenSecret,
Datacenter: datacenter,
QueryOptions: structs.QueryOptions{Token: masterToken},
}
var out structs.ACLTokenResponse
err := msgpackrpc.CallWithCodec(codec, "ACL.TokenRead", &arg, &out)
if err != nil {
return "", err
}
if out.Token == nil {
return "", nil
}
return out.Token.AccessorID, nil
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// retrieveTestToken returns a policy for testing purposes
func retrieveTestToken(codec rpc.ClientCodec, masterToken string, datacenter string, id string) (*structs.ACLTokenResponse, error) {
arg := structs.ACLTokenGetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: datacenter,
TokenID: id,
TokenIDType: structs.ACLTokenAccessor,
QueryOptions: structs.QueryOptions{Token: masterToken},
}
var out structs.ACLTokenResponse
err := msgpackrpc.CallWithCodec(codec, "ACL.TokenRead", &arg, &out)
if err != nil {
return nil, err
}
return &out, nil
}
func deleteTestToken(codec rpc.ClientCodec, masterToken string, datacenter string, tokenAccessor string) error {
arg := structs.ACLTokenDeleteRequest{
Datacenter: datacenter,
TokenID: tokenAccessor,
WriteRequest: structs.WriteRequest{Token: masterToken},
}
var ignored string
err := msgpackrpc.CallWithCodec(codec, "ACL.TokenDelete", &arg, &ignored)
return err
}
func deleteTestPolicy(codec rpc.ClientCodec, masterToken string, datacenter string, policyID string) error {
arg := structs.ACLPolicyDeleteRequest{
Datacenter: datacenter,
PolicyID: policyID,
WriteRequest: structs.WriteRequest{Token: masterToken},
}
var ignored string
err := msgpackrpc.CallWithCodec(codec, "ACL.PolicyDelete", &arg, &ignored)
return err
}
func upsertTestCustomizedPolicy(codec rpc.ClientCodec, masterToken string, datacenter string, policyModificationFn func(policy *structs.ACLPolicy)) (*structs.ACLPolicy, error) {
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// Make sure test policies can't collide
policyUnq, err := uuid.GenerateUUID()
if err != nil {
return nil, err
}
arg := structs.ACLPolicySetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: datacenter,
Policy: structs.ACLPolicy{
Name: fmt.Sprintf("test-policy-%s", policyUnq),
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
},
WriteRequest: structs.WriteRequest{Token: masterToken},
}
if policyModificationFn != nil {
policyModificationFn(&arg.Policy)
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
var out structs.ACLPolicy
err = msgpackrpc.CallWithCodec(codec, "ACL.PolicySet", &arg, &out)
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
if err != nil {
return nil, err
}
if out.ID == "" {
return nil, fmt.Errorf("ID is nil: %v", out)
}
return &out, nil
}
// upsertTestPolicy creates a policy for testing purposes
func upsertTestPolicy(codec rpc.ClientCodec, masterToken string, datacenter string) (*structs.ACLPolicy, error) {
return upsertTestPolicyWithRules(codec, masterToken, datacenter, "")
}
func upsertTestPolicyWithRules(codec rpc.ClientCodec, masterToken string, datacenter string, rules string) (*structs.ACLPolicy, error) {
return upsertTestCustomizedPolicy(codec, masterToken, datacenter, func(policy *structs.ACLPolicy) {
policy.Rules = rules
})
}
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
// retrieveTestPolicy returns a policy for testing purposes
func retrieveTestPolicy(codec rpc.ClientCodec, masterToken string, datacenter string, id string) (*structs.ACLPolicyResponse, error) {
arg := structs.ACLPolicyGetRequest{
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
Datacenter: datacenter,
PolicyID: id,
QueryOptions: structs.QueryOptions{Token: masterToken},
}
var out structs.ACLPolicyResponse
err := msgpackrpc.CallWithCodec(codec, "ACL.PolicyRead", &arg, &out)
if err != nil {
return nil, err
}
return &out, nil
}
func deleteTestRole(codec rpc.ClientCodec, masterToken string, datacenter string, roleID string) error {
arg := structs.ACLRoleDeleteRequest{
Datacenter: datacenter,
RoleID: roleID,
WriteRequest: structs.WriteRequest{Token: masterToken},
}
var ignored string
err := msgpackrpc.CallWithCodec(codec, "ACL.RoleDelete", &arg, &ignored)
return err
}
func deleteTestRoleByName(codec rpc.ClientCodec, masterToken string, datacenter string, roleName string) error {
resp, err := retrieveTestRoleByName(codec, masterToken, datacenter, roleName)
if err != nil {
return err
}
if resp.Role == nil {
return nil
}
return deleteTestRole(codec, masterToken, datacenter, resp.Role.ID)
}
// upsertTestRole creates a role for testing purposes
func upsertTestRole(codec rpc.ClientCodec, masterToken string, datacenter string) (*structs.ACLRole, error) {
return upsertTestCustomizedRole(codec, masterToken, datacenter, nil)
}
func upsertTestCustomizedRole(codec rpc.ClientCodec, masterToken string, datacenter string, modify func(role *structs.ACLRole)) (*structs.ACLRole, error) {
// Make sure test roles can't collide
roleUnq, err := uuid.GenerateUUID()
if err != nil {
return nil, err
}
arg := structs.ACLRoleSetRequest{
Datacenter: datacenter,
Role: structs.ACLRole{
Name: fmt.Sprintf("test-role-%s", roleUnq),
},
WriteRequest: structs.WriteRequest{Token: masterToken},
}
if modify != nil {
modify(&arg.Role)
}
var out structs.ACLRole
err = msgpackrpc.CallWithCodec(codec, "ACL.RoleSet", &arg, &out)
if err != nil {
return nil, err
}
if out.ID == "" {
return nil, fmt.Errorf("ID is nil: %v", out)
}
return &out, nil
}
func retrieveTestRole(codec rpc.ClientCodec, masterToken string, datacenter string, id string) (*structs.ACLRoleResponse, error) {
arg := structs.ACLRoleGetRequest{
Datacenter: datacenter,
RoleID: id,
QueryOptions: structs.QueryOptions{Token: masterToken},
}
var out structs.ACLRoleResponse
err := msgpackrpc.CallWithCodec(codec, "ACL.RoleRead", &arg, &out)
if err != nil {
return nil, err
}
return &out, nil
}
func retrieveTestRoleByName(codec rpc.ClientCodec, masterToken string, datacenter string, name string) (*structs.ACLRoleResponse, error) {
arg := structs.ACLRoleGetRequest{
Datacenter: datacenter,
RoleName: name,
QueryOptions: structs.QueryOptions{Token: masterToken},
}
var out structs.ACLRoleResponse
err := msgpackrpc.CallWithCodec(codec, "ACL.RoleRead", &arg, &out)
if err != nil {
return nil, err
}
return &out, nil
}
func deleteTestAuthMethod(codec rpc.ClientCodec, masterToken string, datacenter string, methodName string) error {
arg := structs.ACLAuthMethodDeleteRequest{
Datacenter: datacenter,
AuthMethodName: methodName,
WriteRequest: structs.WriteRequest{Token: masterToken},
}
var ignored string
err := msgpackrpc.CallWithCodec(codec, "ACL.AuthMethodDelete", &arg, &ignored)
return err
}
func upsertTestAuthMethod(
codec rpc.ClientCodec, masterToken string, datacenter string,
sessionID string,
) (*structs.ACLAuthMethod, error) {
return upsertTestCustomizedAuthMethod(codec, masterToken, datacenter, func(method *structs.ACLAuthMethod) {
method.Config = map[string]interface{}{
"SessionID": sessionID,
}
})
}
func upsertTestCustomizedAuthMethod(
codec rpc.ClientCodec, masterToken string, datacenter string,
modify func(method *structs.ACLAuthMethod),
) (*structs.ACLAuthMethod, error) {
name, err := uuid.GenerateUUID()
if err != nil {
return nil, err
}
req := structs.ACLAuthMethodSetRequest{
Datacenter: datacenter,
AuthMethod: structs.ACLAuthMethod{
Name: "test-method-" + name,
Type: "testing",
},
WriteRequest: structs.WriteRequest{Token: masterToken},
}
if modify != nil {
modify(&req.AuthMethod)
}
var out structs.ACLAuthMethod
err = msgpackrpc.CallWithCodec(codec, "ACL.AuthMethodSet", &req, &out)
if err != nil {
return nil, err
}
return &out, nil
}
func upsertTestKubernetesAuthMethod(
codec rpc.ClientCodec, masterToken string, datacenter string,
caCert, kubeHost, kubeJWT string,
) (*structs.ACLAuthMethod, error) {
name, err := uuid.GenerateUUID()
if err != nil {
return nil, err
}
if kubeHost == "" {
kubeHost = "https://abc:8443"
}
if kubeJWT == "" {
kubeJWT = goodJWT_A
}
req := structs.ACLAuthMethodSetRequest{
Datacenter: datacenter,
AuthMethod: structs.ACLAuthMethod{
Name: "test-method-" + name,
Type: "kubernetes",
Config: map[string]interface{}{
"Host": kubeHost,
"CACert": caCert,
"ServiceAccountJWT": kubeJWT,
},
},
WriteRequest: structs.WriteRequest{Token: masterToken},
}
var out structs.ACLAuthMethod
err = msgpackrpc.CallWithCodec(codec, "ACL.AuthMethodSet", &req, &out)
if err != nil {
return nil, err
}
return &out, nil
}
func retrieveTestAuthMethod(codec rpc.ClientCodec, masterToken string, datacenter string, name string) (*structs.ACLAuthMethodResponse, error) {
arg := structs.ACLAuthMethodGetRequest{
Datacenter: datacenter,
AuthMethodName: name,
QueryOptions: structs.QueryOptions{Token: masterToken},
}
var out structs.ACLAuthMethodResponse
err := msgpackrpc.CallWithCodec(codec, "ACL.AuthMethodRead", &arg, &out)
if err != nil {
return nil, err
}
return &out, nil
}
func deleteTestBindingRule(codec rpc.ClientCodec, masterToken string, datacenter string, ruleID string) error {
arg := structs.ACLBindingRuleDeleteRequest{
Datacenter: datacenter,
BindingRuleID: ruleID,
WriteRequest: structs.WriteRequest{Token: masterToken},
}
var ignored string
err := msgpackrpc.CallWithCodec(codec, "ACL.BindingRuleDelete", &arg, &ignored)
return err
}
func upsertTestBindingRule(
codec rpc.ClientCodec,
masterToken string,
datacenter string,
methodName string,
selector string,
bindType string,
bindName string,
) (*structs.ACLBindingRule, error) {
return upsertTestCustomizedBindingRule(codec, masterToken, datacenter, func(rule *structs.ACLBindingRule) {
rule.AuthMethod = methodName
rule.BindType = bindType
rule.BindName = bindName
rule.Selector = selector
})
}
func upsertTestCustomizedBindingRule(codec rpc.ClientCodec, masterToken string, datacenter string, modify func(rule *structs.ACLBindingRule)) (*structs.ACLBindingRule, error) {
req := structs.ACLBindingRuleSetRequest{
Datacenter: datacenter,
BindingRule: structs.ACLBindingRule{},
WriteRequest: structs.WriteRequest{Token: masterToken},
}
if modify != nil {
modify(&req.BindingRule)
}
var out structs.ACLBindingRule
err := msgpackrpc.CallWithCodec(codec, "ACL.BindingRuleSet", &req, &out)
if err != nil {
return nil, err
}
return &out, nil
}
func retrieveTestBindingRule(codec rpc.ClientCodec, masterToken string, datacenter string, ruleID string) (*structs.ACLBindingRuleResponse, error) {
arg := structs.ACLBindingRuleGetRequest{
Datacenter: datacenter,
BindingRuleID: ruleID,
QueryOptions: structs.QueryOptions{Token: masterToken},
}
var out structs.ACLBindingRuleResponse
err := msgpackrpc.CallWithCodec(codec, "ACL.BindingRuleRead", &arg, &out)
if err != nil {
return nil, err
}
return &out, nil
}
func requireTimeEquals(t *testing.T, expect, got *time.Time) {
t.Helper()
if expect == nil && got == nil {
return
} else if expect == nil && got != nil {
t.Fatalf("expected=NIL != got=%q", *got)
} else if expect != nil && got == nil {
t.Fatalf("expected=%q != got=NIL", *expect)
} else if !expect.Equal(*got) {
t.Fatalf("expected=%q != got=%q", *expect, *got)
}
}
func requireErrorContains(t *testing.T, err error, expectedErrorMessage string) {
t.Helper()
if err == nil {
t.Fatal("An error is expected but got nil.")
}
if !strings.Contains(err.Error(), expectedErrorMessage) {
t.Fatalf("unexpected error: %v", err)
}
}
// 'default/admin'
const goodJWT_A = "eyJhbGciOiJSUzI1NiIsImtpZCI6IiJ9.eyJpc3MiOiJrdWJlcm5ldGVzL3NlcnZpY2VhY2NvdW50Iiwia3ViZXJuZXRlcy5pby9zZXJ2aWNlYWNjb3VudC9uYW1lc3BhY2UiOiJkZWZhdWx0Iiwia3ViZXJuZXRlcy5pby9zZXJ2aWNlYWNjb3VudC9zZWNyZXQubmFtZSI6ImFkbWluLXRva2VuLXFsejQyIiwia3ViZXJuZXRlcy5pby9zZXJ2aWNlYWNjb3VudC9zZXJ2aWNlLWFjY291bnQubmFtZSI6ImFkbWluIiwia3ViZXJuZXRlcy5pby9zZXJ2aWNlYWNjb3VudC9zZXJ2aWNlLWFjY291bnQudWlkIjoiNzM4YmMyNTEtNjUzMi0xMWU5LWI2N2YtNDhlNmM4YjhlY2I1Iiwic3ViIjoic3lzdGVtOnNlcnZpY2VhY2NvdW50OmRlZmF1bHQ6YWRtaW4ifQ.ixMlnWrAG7NVuTTKu8cdcYfM7gweS3jlKaEsIBNGOVEjPE7rtXtgMkAwjQTdYR08_0QBjkgzy5fQC5ZNyglSwONJ-bPaXGvhoH1cTnRi1dz9H_63CfqOCvQP1sbdkMeRxNTGVAyWZT76rXoCUIfHP4LY2I8aab0KN9FTIcgZRF0XPTtT70UwGIrSmRpxW38zjiy2ymWL01cc5VWGhJqVysmWmYk3wNp0h5N57H_MOrz4apQR4pKaamzskzjLxO55gpbmZFC76qWuUdexAR7DT2fpbHLOw90atN_NlLMY-VrXyW3-Ei5EhYaVreMB9PSpKwkrA4jULITohV-sxpa1LA"
// 'default/demo'
const goodJWT_B = "eyJhbGciOiJSUzI1NiIsImtpZCI6IiJ9.eyJpc3MiOiJrdWJlcm5ldGVzL3NlcnZpY2VhY2NvdW50Iiwia3ViZXJuZXRlcy5pby9zZXJ2aWNlYWNjb3VudC9uYW1lc3BhY2UiOiJkZWZhdWx0Iiwia3ViZXJuZXRlcy5pby9zZXJ2aWNlYWNjb3VudC9zZWNyZXQubmFtZSI6ImRlbW8tdG9rZW4ta21iOW4iLCJrdWJlcm5ldGVzLmlvL3NlcnZpY2VhY2NvdW50L3NlcnZpY2UtYWNjb3VudC5uYW1lIjoiZGVtbyIsImt1YmVybmV0ZXMuaW8vc2VydmljZWFjY291bnQvc2VydmljZS1hY2NvdW50LnVpZCI6Ijc2MDkxYWY0LTRiNTYtMTFlOS1hYzRiLTcwOGIxMTgwMWNiZSIsInN1YiI6InN5c3RlbTpzZXJ2aWNlYWNjb3VudDpkZWZhdWx0OmRlbW8ifQ.ZiAHjijBAOsKdum0Aix6lgtkLkGo9_Tu87dWQ5Zfwnn3r2FejEWDAnftTft1MqqnMzivZ9Wyyki5ZjQRmTAtnMPJuHC-iivqY4Wh4S6QWCJ1SivBv5tMZR79t5t8mE7R1-OHwst46spru1pps9wt9jsA04d3LpV0eeKYgdPTVaQKklxTm397kIMUugA6yINIBQ3Rh8eQqBgNwEmL4iqyYubzHLVkGkoP9MJikFI05vfRiHtYr-piXz6JFDzXMQj9rW6xtMmrBSn79ChbyvC5nz-Nj2rJPnHsb_0rDUbmXY5PpnMhBpdSH-CbZ4j8jsiib6DtaGJhVZeEQ1GjsFAZwQ"