open-consul/agent/consul/coordinate_endpoint_test.go

622 lines
17 KiB
Go
Raw Normal View History

2015-04-09 20:23:14 +00:00
package consul
import (
2015-05-14 01:22:34 +00:00
"fmt"
"math"
2015-04-09 20:23:14 +00:00
"math/rand"
2017-10-31 20:34:49 +00:00
"net/rpc"
2015-04-09 20:23:14 +00:00
"os"
"strings"
2015-04-09 20:23:14 +00:00
"testing"
2015-04-13 20:45:42 +00:00
"time"
2015-04-09 20:23:14 +00:00
"github.com/hashicorp/consul/acl"
"github.com/hashicorp/consul/agent/structs"
"github.com/hashicorp/consul/lib"
"github.com/hashicorp/consul/sdk/testutil/retry"
"github.com/hashicorp/consul/testrpc"
msgpackrpc "github.com/hashicorp/net-rpc-msgpackrpc"
2015-04-09 20:23:14 +00:00
"github.com/hashicorp/serf/coordinate"
"github.com/stretchr/testify/require"
2015-04-09 20:23:14 +00:00
)
// generateRandomCoordinate creates a random coordinate. This mucks with the
// underlying structure directly, so it's not really useful for any particular
// position in the network, but it's a good payload to send through to make
// sure things come out the other side or get stored correctly.
func generateRandomCoordinate() *coordinate.Coordinate {
2015-04-09 20:23:14 +00:00
config := coordinate.DefaultConfig()
coord := coordinate.NewCoordinate(config)
for i := range coord.Vec {
coord.Vec[i] = rand.NormFloat64()
2015-04-13 20:45:42 +00:00
}
coord.Error = rand.NormFloat64()
coord.Adjustment = rand.NormFloat64()
return coord
2015-04-09 20:23:14 +00:00
}
2015-05-14 02:09:58 +00:00
func TestCoordinate_Update(t *testing.T) {
t.Parallel()
dir1, s1 := testServerWithConfig(t, func(c *Config) {
c.CoordinateUpdatePeriod = 500 * time.Millisecond
c.CoordinateUpdateBatchSize = 5
c.CoordinateUpdateMaxBatches = 2
})
defer os.RemoveAll(dir1)
2015-04-09 20:23:14 +00:00
defer s1.Shutdown()
2015-05-14 01:22:34 +00:00
2015-10-15 23:07:16 +00:00
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForTestAgent(t, s1.RPC, "dc1")
2015-04-09 20:23:14 +00:00
// Register some nodes.
nodes := []string{"node1", "node2"}
if err := registerNodes(nodes, codec, ""); err != nil {
2017-10-31 20:34:49 +00:00
t.Fatal(err)
}
// Send an update for the first node.
2015-05-08 08:31:34 +00:00
arg1 := structs.CoordinateUpdateRequest{
2015-04-18 21:05:29 +00:00
Datacenter: "dc1",
Node: "node1",
Coord: generateRandomCoordinate(),
2015-04-09 20:23:14 +00:00
}
var out struct{}
2015-10-15 23:07:16 +00:00
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg1, &out); err != nil {
t.Fatalf("err: %v", err)
}
2015-04-09 20:23:14 +00:00
// Send an update for the second node.
2015-05-08 08:31:34 +00:00
arg2 := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "node2",
Coord: generateRandomCoordinate(),
2015-05-08 08:31:34 +00:00
}
2015-10-15 23:07:16 +00:00
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg2, &out); err != nil {
2015-04-09 20:23:14 +00:00
t.Fatalf("err: %v", err)
}
// Make sure the updates did not yet apply because the update period
// hasn't expired.
2015-04-09 20:23:14 +00:00
state := s1.fsm.State()
_, c, err := state.Coordinate("node1", nil)
2015-04-09 20:23:14 +00:00
if err != nil {
t.Fatalf("err: %v", err)
}
require.Equal(t, lib.CoordinateSet{}, c)
_, c, err = state.Coordinate("node2", nil)
if err != nil {
2015-05-08 08:31:34 +00:00
t.Fatalf("err: %v", err)
}
require.Equal(t, lib.CoordinateSet{}, c)
// Send another update for the second node. It should take precedence
// since there will be two updates in the same batch.
arg2.Coord = generateRandomCoordinate()
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg2, &out); err != nil {
t.Fatalf("err: %v", err)
}
// Wait a while and the updates should get picked up.
time.Sleep(3 * s1.config.CoordinateUpdatePeriod)
_, c, err = state.Coordinate("node1", nil)
2015-05-08 08:31:34 +00:00
if err != nil {
t.Fatalf("err: %v", err)
}
expected := lib.CoordinateSet{
"": arg1.Coord,
2015-05-08 08:31:34 +00:00
}
require.Equal(t, expected, c)
_, c, err = state.Coordinate("node2", nil)
2015-05-08 08:31:34 +00:00
if err != nil {
t.Fatalf("err: %v", err)
}
expected = lib.CoordinateSet{
"": arg2.Coord,
2015-05-08 08:31:34 +00:00
}
require.Equal(t, expected, c)
// Register a bunch of additional nodes.
spamLen := s1.config.CoordinateUpdateBatchSize*s1.config.CoordinateUpdateMaxBatches + 1
for i := 0; i < spamLen; i++ {
req := structs.RegisterRequest{
Datacenter: "dc1",
Node: fmt.Sprintf("bogusnode%d", i),
Address: "127.0.0.1",
}
var reply struct{}
if err := msgpackrpc.CallWithCodec(codec, "Catalog.Register", &req, &reply); err != nil {
t.Fatalf("err: %v", err)
}
}
// Now spam some coordinate updates and make sure it starts throwing
// them away if they exceed the batch allowance. Note we have to make
// unique names since these are held in map by node name.
for i := 0; i < spamLen; i++ {
arg1.Node = fmt.Sprintf("bogusnode%d", i)
arg1.Coord = generateRandomCoordinate()
2015-10-15 23:07:16 +00:00
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg1, &out); err != nil {
t.Fatalf("err: %v", err)
}
2015-04-09 20:23:14 +00:00
}
// Wait a little while for the batch routine to run, then make sure
// exactly one of the updates got dropped (we won't know which one).
time.Sleep(3 * s1.config.CoordinateUpdatePeriod)
numDropped := 0
for i := 0; i < spamLen; i++ {
_, c, err = state.Coordinate(fmt.Sprintf("bogusnode%d", i), nil)
if err != nil {
t.Fatalf("err: %v", err)
}
if len(c) == 0 {
numDropped++
}
}
if numDropped != 1 {
t.Fatalf("wrong number of coordinates dropped, %d != 1", numDropped)
}
// Send a coordinate with a NaN to make sure that we don't absorb that
// into the database.
arg2.Coord.Vec[0] = math.NaN()
err = msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg2, &out)
if err == nil || !strings.Contains(err.Error(), "invalid coordinate") {
t.Fatalf("should have failed with an error, got %v", err)
}
// Finally, send a coordinate with the wrong dimensionality to make sure
// there are no panics, and that it gets rejected.
arg2.Coord.Vec = make([]float64, 2*len(arg2.Coord.Vec))
2015-10-15 23:07:16 +00:00
err = msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg2, &out)
if err == nil || !strings.Contains(err.Error(), "incompatible coordinate") {
t.Fatalf("should have failed with an error, got %v", err)
}
2015-04-29 01:47:41 +00:00
}
func TestCoordinate_Update_ACLDeny(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"
c.ACLDefaultPolicy = "deny"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1", testrpc.WithToken("root"))
// Register some nodes.
nodes := []string{"node1", "node2"}
if err := registerNodes(nodes, codec, "root"); err != nil {
2017-10-31 20:34:49 +00:00
t.Fatal(err)
}
// Send an update for the first node.
// don't have version 8 ACLs enforced yet.
req := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "node1",
Coord: generateRandomCoordinate(),
}
var out struct{}
err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &req, &out)
if !acl.IsErrPermissionDenied(err) {
t.Fatalf("err: %v", err)
}
// Create an ACL that can write to the node.
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,
Rules: `
node "node1" {
policy = "write"
}
`,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
var id string
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &arg, &id); err != nil {
t.Fatalf("err: %v", err)
}
// With the token, it should now go through.
req.Token = id
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &req, &out); err != nil {
t.Fatalf("err: %v", err)
}
// But it should be blocked for the other node.
req.Node = "node2"
err = msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &req, &out)
if !acl.IsErrPermissionDenied(err) {
t.Fatalf("err: %v", err)
}
}
func TestCoordinate_ListDatacenters(t *testing.T) {
t.Parallel()
dir1, s1 := testServer(t)
defer os.RemoveAll(dir1)
defer s1.Shutdown()
2015-10-15 23:07:16 +00:00
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
// It's super hard to force the Serfs into a known configuration of
// coordinates, so the best we can do is make sure our own DC shows
// up in the list with the proper coordinates. The guts of the algorithm
// are extensively tested in rtt_test.go using a mock database.
var out []structs.DatacenterMap
2015-10-15 23:07:16 +00:00
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.ListDatacenters", struct{}{}, &out); err != nil {
t.Fatalf("err: %v", err)
}
if len(out) != 1 ||
out[0].Datacenter != "dc1" ||
len(out[0].Coordinates) != 1 ||
out[0].Coordinates[0].Node != s1.config.NodeName {
t.Fatalf("bad: %v", out)
}
c, err := s1.serfWAN.GetCoordinate()
if err != nil {
t.Fatalf("bad: %v", err)
}
require.Equal(t, out[0].Coordinates[0].Coord, c)
}
func TestCoordinate_ListNodes(t *testing.T) {
t.Parallel()
dir1, s1 := testServer(t)
defer os.RemoveAll(dir1)
defer s1.Shutdown()
2015-10-15 23:07:16 +00:00
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1")
// Register some nodes.
nodes := []string{"foo", "bar", "baz"}
if err := registerNodes(nodes, codec, ""); err != nil {
2017-10-31 20:34:49 +00:00
t.Fatal(err)
}
// Send coordinate updates for a few nodes.
arg1 := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "foo",
Coord: generateRandomCoordinate(),
}
var out struct{}
2015-10-15 23:07:16 +00:00
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg1, &out); err != nil {
t.Fatalf("err: %v", err)
}
arg2 := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "bar",
Coord: generateRandomCoordinate(),
}
2015-10-15 23:07:16 +00:00
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg2, &out); err != nil {
t.Fatalf("err: %v", err)
}
arg3 := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "baz",
Coord: generateRandomCoordinate(),
}
2015-10-15 23:07:16 +00:00
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg3, &out); err != nil {
t.Fatalf("err: %v", err)
}
// Now query back for all the nodes.
retry.Run(t, func(r *retry.R) {
arg := structs.DCSpecificRequest{
Datacenter: "dc1",
}
resp := structs.IndexedCoordinates{}
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.ListNodes", &arg, &resp); err != nil {
r.Fatalf("err: %v", err)
}
if len(resp.Coordinates) != 3 ||
resp.Coordinates[0].Node != "bar" ||
resp.Coordinates[1].Node != "baz" ||
resp.Coordinates[2].Node != "foo" {
r.Fatalf("bad: %v", resp.Coordinates)
}
require.Equal(r, arg2.Coord, resp.Coordinates[0].Coord) // bar
require.Equal(r, arg3.Coord, resp.Coordinates[1].Coord) // baz
require.Equal(r, arg1.Coord, resp.Coordinates[2].Coord) // foo
})
}
func TestCoordinate_ListNodes_ACLFilter(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"
c.ACLDefaultPolicy = "deny"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1", testrpc.WithToken("root"))
// Register some nodes.
nodes := []string{"foo", "bar", "baz"}
for _, node := range nodes {
req := structs.RegisterRequest{
Datacenter: "dc1",
Node: node,
Address: "127.0.0.1",
WriteRequest: structs.WriteRequest{
Token: "root",
},
}
var reply struct{}
if err := msgpackrpc.CallWithCodec(codec, "Catalog.Register", &req, &reply); err != nil {
t.Fatalf("err: %v", err)
}
}
// Send coordinate updates for a few nodes.
arg1 := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "foo",
Coord: generateRandomCoordinate(),
WriteRequest: structs.WriteRequest{
Token: "root",
},
}
var out struct{}
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg1, &out); err != nil {
t.Fatalf("err: %v", err)
}
arg2 := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "bar",
Coord: generateRandomCoordinate(),
WriteRequest: structs.WriteRequest{
Token: "root",
},
}
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg2, &out); err != nil {
t.Fatalf("err: %v", err)
}
arg3 := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "baz",
Coord: generateRandomCoordinate(),
WriteRequest: structs.WriteRequest{
Token: "root",
},
}
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg3, &out); err != nil {
t.Fatalf("err: %v", err)
}
// Wait for all the coordinate updates to apply.
retry.Run(t, func(r *retry.R) {
arg := structs.DCSpecificRequest{
Datacenter: "dc1",
QueryOptions: structs.QueryOptions{Token: "root"},
}
resp := structs.IndexedCoordinates{}
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.ListNodes", &arg, &resp); err != nil {
r.Fatalf("err: %v", err)
}
if got, want := len(resp.Coordinates), 3; got != want {
r.Fatalf("got %d coordinates want %d", got, want)
}
})
// Now that we've waited for the batch processing to ingest the
// coordinates we can do the rest of the requests without the loop.
arg := structs.DCSpecificRequest{
Datacenter: "dc1",
}
resp := structs.IndexedCoordinates{}
2015-10-15 23:07:16 +00:00
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.ListNodes", &arg, &resp); err != nil {
t.Fatalf("err: %v", err)
}
if len(resp.Coordinates) != 0 {
t.Fatalf("bad: %#v", resp.Coordinates)
}
// Create an ACL that can read one of the nodes.
var id string
{
req := 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,
Rules: `
node "foo" {
policy = "read"
}
`,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &req, &id); err != nil {
t.Fatalf("err: %v", err)
}
}
// With the token, it should now go through.
arg.Token = id
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.ListNodes", &arg, &resp); err != nil {
t.Fatalf("err: %v", err)
}
if len(resp.Coordinates) != 1 || resp.Coordinates[0].Node != "foo" {
t.Fatalf("bad: %#v", resp.Coordinates)
}
}
func TestCoordinate_Node(t *testing.T) {
t.Parallel()
dir1, s1 := testServer(t)
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForTestAgent(t, s1.RPC, "dc1")
// Register some nodes.
nodes := []string{"foo", "bar"}
if err := registerNodes(nodes, codec, ""); err != nil {
2017-10-31 20:34:49 +00:00
t.Fatal(err)
}
// Send coordinate updates for each node.
arg1 := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "foo",
Coord: generateRandomCoordinate(),
}
var out struct{}
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg1, &out); err != nil {
t.Fatalf("err: %v", err)
}
arg2 := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "bar",
Coord: generateRandomCoordinate(),
}
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &arg2, &out); err != nil {
t.Fatalf("err: %v", err)
}
// Now query back for a specific node (make sure we only get coordinates for foo).
retry.Run(t, func(r *retry.R) {
arg := structs.NodeSpecificRequest{
Node: "foo",
Datacenter: "dc1",
}
resp := structs.IndexedCoordinates{}
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Node", &arg, &resp); err != nil {
r.Fatalf("err: %v", err)
}
if len(resp.Coordinates) != 1 ||
resp.Coordinates[0].Node != "foo" {
r.Fatalf("bad: %v", resp.Coordinates)
}
require.Equal(r, arg1.Coord, resp.Coordinates[0].Coord) // foo
})
}
func TestCoordinate_Node_ACLDeny(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"
c.ACLDefaultPolicy = "deny"
})
defer os.RemoveAll(dir1)
defer s1.Shutdown()
codec := rpcClient(t, s1)
defer codec.Close()
testrpc.WaitForLeader(t, s1.RPC, "dc1", testrpc.WithToken("root"))
// Register some nodes.
nodes := []string{"node1", "node2"}
if err := registerNodes(nodes, codec, "root"); err != nil {
2017-10-31 20:34:49 +00:00
t.Fatal(err)
}
coord := generateRandomCoordinate()
req := structs.CoordinateUpdateRequest{
Datacenter: "dc1",
Node: "node1",
Coord: coord,
WriteRequest: structs.WriteRequest{Token: "root"},
}
var out struct{}
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Update", &req, &out); err != nil {
t.Fatalf("err: %v", err)
}
// Try a read for the first node. This should fail without a token.
arg := structs.NodeSpecificRequest{
Node: "node1",
Datacenter: "dc1",
}
resp := structs.IndexedCoordinates{}
err := msgpackrpc.CallWithCodec(codec, "Coordinate.Node", &arg, &resp)
if !acl.IsErrPermissionDenied(err) {
t.Fatalf("err: %v", err)
}
// Create an ACL that can read from the node.
aclReq := 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,
Rules: `
node "node1" {
policy = "read"
}
`,
},
WriteRequest: structs.WriteRequest{Token: "root"},
}
var id string
if err := msgpackrpc.CallWithCodec(codec, "ACL.Apply", &aclReq, &id); err != nil {
t.Fatalf("err: %v", err)
}
// With the token, it should now go through.
arg.Token = id
if err := msgpackrpc.CallWithCodec(codec, "Coordinate.Node", &arg, &resp); err != nil {
t.Fatalf("err: %v", err)
}
// But it should be blocked for the other node.
arg.Node = "node2"
err = msgpackrpc.CallWithCodec(codec, "Coordinate.Node", &arg, &resp)
if !acl.IsErrPermissionDenied(err) {
t.Fatalf("err: %v", err)
}
}
2017-10-31 20:34:49 +00:00
func registerNodes(nodes []string, codec rpc.ClientCodec, token string) error {
2017-10-31 20:34:49 +00:00
for _, node := range nodes {
req := structs.RegisterRequest{
Datacenter: "dc1",
Node: node,
Address: "127.0.0.1",
WriteRequest: structs.WriteRequest{Token: token},
2017-10-31 20:34:49 +00:00
}
var reply struct{}
if err := msgpackrpc.CallWithCodec(codec, "Catalog.Register", &req, &reply); err != nil {
return err
}
}
return nil
}