open-consul/api/acl.go

1113 lines
29 KiB
Go
Raw Normal View History

package api
2017-04-24 21:39:50 +00:00
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"
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
"io/ioutil"
"net/url"
2017-04-24 21:39:50 +00:00
"time"
"github.com/mitchellh/mapstructure"
2017-04-24 21:39:50 +00:00
)
const (
// ACLClientType is the client type token
ACLClientType = "client"
// ACLManagementType is the management type token
ACLManagementType = "management"
)
type ACLLink struct {
ID string
Name string
}
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 ACLTokenPolicyLink = ACLLink
type ACLTokenRoleLink = ACLLink
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
// ACLToken represents an ACL Token
type ACLToken struct {
CreateIndex uint64
ModifyIndex uint64
AccessorID string
SecretID string
Description string
Policies []*ACLTokenPolicyLink `json:",omitempty"`
Roles []*ACLTokenRoleLink `json:",omitempty"`
ServiceIdentities []*ACLServiceIdentity `json:",omitempty"`
Local bool
ExpirationTTL time.Duration `json:",omitempty"`
ExpirationTime *time.Time `json:",omitempty"`
CreateTime time.Time `json:",omitempty"`
Hash []byte `json:",omitempty"`
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
// DEPRECATED (ACL-Legacy-Compat)
// Rules will only be present for legacy tokens returned via the new APIs
Rules string `json:",omitempty"`
}
type ACLTokenListEntry struct {
CreateIndex uint64
ModifyIndex uint64
AccessorID string
Description string
Policies []*ACLTokenPolicyLink `json:",omitempty"`
Roles []*ACLTokenRoleLink `json:",omitempty"`
ServiceIdentities []*ACLServiceIdentity `json:",omitempty"`
Local bool
ExpirationTime *time.Time `json:",omitempty"`
CreateTime time.Time
Hash []byte
Legacy bool
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
}
// ACLEntry is used to represent a legacy ACL token
// The legacy tokens are deprecated.
type ACLEntry struct {
CreateIndex uint64
ModifyIndex uint64
ID string
Name string
Type string
Rules string
}
// ACLReplicationStatus is used to represent the status of ACL replication.
2017-04-25 23:32:09 +00:00
type ACLReplicationStatus struct {
Enabled bool
Running bool
SourceDatacenter string
ReplicationType string
ReplicatedIndex uint64
ReplicatedRoleIndex uint64
ReplicatedTokenIndex uint64
LastSuccess time.Time
LastError time.Time
2017-04-24 21:39:50 +00:00
}
// ACLServiceIdentity represents a high-level grant of all necessary privileges
// to assume the identity of the named Service in the Catalog and within
// Connect.
type ACLServiceIdentity struct {
ServiceName string
Datacenters []string `json:",omitempty"`
}
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
// ACLPolicy represents an ACL Policy.
type ACLPolicy struct {
ID string
Name string
Description string
Rules string
Datacenters []string
Hash []byte
CreateIndex uint64
ModifyIndex uint64
}
type ACLPolicyListEntry struct {
ID string
Name string
Description string
Datacenters []string
Hash []byte
CreateIndex uint64
ModifyIndex uint64
}
type ACLRolePolicyLink = ACLLink
// ACLRole represents an ACL Role.
type ACLRole struct {
ID string
Name string
Description string
Policies []*ACLRolePolicyLink `json:",omitempty"`
ServiceIdentities []*ACLServiceIdentity `json:",omitempty"`
Hash []byte
CreateIndex uint64
ModifyIndex uint64
}
// BindingRuleBindType is the type of binding rule mechanism used.
type BindingRuleBindType string
const (
// BindingRuleBindTypeService binds to a service identity with the given name.
BindingRuleBindTypeService BindingRuleBindType = "service"
// BindingRuleBindTypeRole binds to pre-existing roles with the given name.
BindingRuleBindTypeRole BindingRuleBindType = "role"
)
type ACLBindingRule struct {
ID string
Description string
AuthMethod string
Selector string
BindType BindingRuleBindType
BindName string
CreateIndex uint64
ModifyIndex uint64
}
type ACLAuthMethod struct {
Name string
Type string
Description string
// Configuration is arbitrary configuration for the auth method. This
// should only contain primitive values and containers (such as lists and
// maps).
Config map[string]interface{}
CreateIndex uint64
ModifyIndex uint64
}
type ACLAuthMethodListEntry struct {
Name string
Type string
Description string
CreateIndex uint64
ModifyIndex uint64
}
// ParseKubernetesAuthMethodConfig takes a raw config map and returns a parsed
// KubernetesAuthMethodConfig.
func ParseKubernetesAuthMethodConfig(raw map[string]interface{}) (*KubernetesAuthMethodConfig, error) {
var config KubernetesAuthMethodConfig
decodeConf := &mapstructure.DecoderConfig{
Result: &config,
WeaklyTypedInput: true,
}
decoder, err := mapstructure.NewDecoder(decodeConf)
if err != nil {
return nil, err
}
if err := decoder.Decode(raw); err != nil {
return nil, fmt.Errorf("error decoding config: %s", err)
}
return &config, nil
}
// KubernetesAuthMethodConfig is the config for the built-in Consul auth method
// for Kubernetes.
type KubernetesAuthMethodConfig struct {
Host string `json:",omitempty"`
CACert string `json:",omitempty"`
ServiceAccountJWT string `json:",omitempty"`
}
// RenderToConfig converts this into a map[string]interface{} suitable for use
// in the ACLAuthMethod.Config field.
func (c *KubernetesAuthMethodConfig) RenderToConfig() map[string]interface{} {
return map[string]interface{}{
"Host": c.Host,
"CACert": c.CACert,
"ServiceAccountJWT": c.ServiceAccountJWT,
}
}
type ACLLoginParams struct {
AuthMethod string
BearerToken string
Meta map[string]string `json:",omitempty"`
}
// ACL can be used to query the ACL endpoints
type ACL struct {
c *Client
}
// ACL returns a handle to the ACL endpoints
func (c *Client) ACL() *ACL {
return &ACL{c}
}
// Bootstrap is used to perform a one-time ACL bootstrap operation on a cluster
// to get the first management 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 (a *ACL) Bootstrap() (*ACLToken, *WriteMeta, error) {
r := a.c.newRequest("PUT", "/v1/acl/bootstrap")
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != 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
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
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 ACLToken
if err := decodeBody(resp, &out); err != 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
return nil, nil, 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
return &out, wm, nil
}
// Create is used to generate a new token with the given parameters
//
// Deprecated: Use TokenCreate instead.
func (a *ACL) Create(acl *ACLEntry, q *WriteOptions) (string, *WriteMeta, error) {
r := a.c.newRequest("PUT", "/v1/acl/create")
r.setWriteOptions(q)
r.obj = acl
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return "", nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out struct{ ID string }
if err := decodeBody(resp, &out); err != nil {
return "", nil, err
}
return out.ID, wm, nil
}
// Update is used to update the rules of an existing token
//
// Deprecated: Use TokenUpdate instead.
func (a *ACL) Update(acl *ACLEntry, q *WriteOptions) (*WriteMeta, error) {
r := a.c.newRequest("PUT", "/v1/acl/update")
r.setWriteOptions(q)
r.obj = acl
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
return wm, nil
}
// Destroy is used to destroy a given ACL token ID
//
// Deprecated: Use TokenDelete instead.
func (a *ACL) Destroy(id string, q *WriteOptions) (*WriteMeta, error) {
r := a.c.newRequest("PUT", "/v1/acl/destroy/"+id)
r.setWriteOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, err
}
resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
return wm, nil
}
// Clone is used to return a new token cloned from an existing one
//
// Deprecated: Use TokenClone instead.
func (a *ACL) Clone(id string, q *WriteOptions) (string, *WriteMeta, error) {
r := a.c.newRequest("PUT", "/v1/acl/clone/"+id)
r.setWriteOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return "", nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out struct{ ID string }
if err := decodeBody(resp, &out); err != nil {
return "", nil, err
}
return out.ID, wm, nil
}
// Info is used to query for information about an ACL token
//
// Deprecated: Use TokenRead instead.
func (a *ACL) Info(id string, q *QueryOptions) (*ACLEntry, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/info/"+id)
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var entries []*ACLEntry
if err := decodeBody(resp, &entries); err != nil {
return nil, nil, err
}
if len(entries) > 0 {
return entries[0], qm, nil
}
return nil, qm, nil
}
// List is used to get all the ACL tokens
//
// Deprecated: Use TokenList instead.
func (a *ACL) List(q *QueryOptions) ([]*ACLEntry, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/list")
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var entries []*ACLEntry
if err := decodeBody(resp, &entries); err != nil {
return nil, nil, err
}
return entries, qm, nil
}
2017-04-24 21:39:50 +00:00
// Replication returns the status of the ACL replication process in the datacenter
func (a *ACL) Replication(q *QueryOptions) (*ACLReplicationStatus, *QueryMeta, error) {
2017-04-24 21:39:50 +00:00
r := a.c.newRequest("GET", "/v1/acl/replication")
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var entries *ACLReplicationStatus
2017-04-24 21:39:50 +00:00
if err := decodeBody(resp, &entries); err != nil {
return nil, nil, err
}
return entries, qm, 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
2019-04-30 15:45:36 +00:00
// TokenCreate creates a new ACL token. If either the AccessorID or SecretID fields
// of the ACLToken structure are empty they will be filled in by Consul.
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 (a *ACL) TokenCreate(token *ACLToken, q *WriteOptions) (*ACLToken, *WriteMeta, error) {
r := a.c.newRequest("PUT", "/v1/acl/token")
r.setWriteOptions(q)
r.obj = token
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLToken
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// TokenUpdate updates a token in place without modifying its AccessorID or SecretID. A valid
// AccessorID must be set in the ACLToken structure passed to this function but the SecretID may
// be omitted and will be filled in by Consul with its existing value.
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 (a *ACL) TokenUpdate(token *ACLToken, q *WriteOptions) (*ACLToken, *WriteMeta, error) {
if token.AccessorID == "" {
return nil, nil, fmt.Errorf("Must specify an AccessorID for Token Updating")
}
r := a.c.newRequest("PUT", "/v1/acl/token/"+token.AccessorID)
r.setWriteOptions(q)
r.obj = token
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLToken
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// TokenClone will create a new token with the same policies and locality as the original
// token but will have its own auto-generated AccessorID and SecretID as well having the
// description passed to this function. The tokenID parameter must be a valid Accessor ID
// of an existing 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 (a *ACL) TokenClone(tokenID string, description string, q *WriteOptions) (*ACLToken, *WriteMeta, error) {
if tokenID == "" {
return nil, nil, fmt.Errorf("Must specify a tokenID for Token Cloning")
}
r := a.c.newRequest("PUT", "/v1/acl/token/"+tokenID+"/clone")
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
r.setWriteOptions(q)
r.obj = struct{ Description string }{description}
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLToken
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// TokenDelete removes a single ACL token. The tokenID parameter must be a valid
// Accessor ID of an existing 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 (a *ACL) TokenDelete(tokenID string, q *WriteOptions) (*WriteMeta, error) {
r := a.c.newRequest("DELETE", "/v1/acl/token/"+tokenID)
r.setWriteOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, err
}
resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
return wm, nil
}
// TokenRead retrieves the full token details. The tokenID parameter must be a valid
// Accessor ID of an existing 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 (a *ACL) TokenRead(tokenID string, q *QueryOptions) (*ACLToken, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/token/"+tokenID)
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var out ACLToken
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, qm, nil
}
// TokenReadSelf retrieves the full token details of the token currently
// assigned to the API Client. In this manner its possible to read a token
// by its Secret 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 (a *ACL) TokenReadSelf(q *QueryOptions) (*ACLToken, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/token/self")
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var out ACLToken
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, qm, nil
}
// TokenList lists all tokens. The listing does not contain any SecretIDs as those
// may only be retrieved by a call to TokenRead.
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 (a *ACL) TokenList(q *QueryOptions) ([]*ACLTokenListEntry, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/tokens")
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var entries []*ACLTokenListEntry
if err := decodeBody(resp, &entries); err != nil {
return nil, nil, err
}
return entries, qm, nil
}
// PolicyCreate will create a new policy. It is not allowed for the policy parameters
// ID field to be set as this will be generated by Consul while processing the request.
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 (a *ACL) PolicyCreate(policy *ACLPolicy, q *WriteOptions) (*ACLPolicy, *WriteMeta, error) {
if policy.ID != "" {
return nil, nil, fmt.Errorf("Cannot specify an ID in Policy Creation")
}
r := a.c.newRequest("PUT", "/v1/acl/policy")
r.setWriteOptions(q)
r.obj = policy
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLPolicy
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// PolicyUpdate updates a policy. The ID field of the policy parameter must be set to an
// existing 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
func (a *ACL) PolicyUpdate(policy *ACLPolicy, q *WriteOptions) (*ACLPolicy, *WriteMeta, error) {
if policy.ID == "" {
return nil, nil, fmt.Errorf("Must specify an ID in Policy Update")
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
}
r := a.c.newRequest("PUT", "/v1/acl/policy/"+policy.ID)
r.setWriteOptions(q)
r.obj = policy
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLPolicy
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// PolicyDelete deletes a policy given its 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 (a *ACL) PolicyDelete(policyID string, q *WriteOptions) (*WriteMeta, error) {
r := a.c.newRequest("DELETE", "/v1/acl/policy/"+policyID)
r.setWriteOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, err
}
resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
return wm, nil
}
// PolicyRead retrieves the policy details including the rule set.
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 (a *ACL) PolicyRead(policyID string, q *QueryOptions) (*ACLPolicy, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/policy/"+policyID)
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var out ACLPolicy
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, qm, nil
}
// PolicyList retrieves a listing of all policies. The listing does not include the
// rules for any policy as those should be retrieved by subsequent calls to PolicyRead.
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 (a *ACL) PolicyList(q *QueryOptions) ([]*ACLPolicyListEntry, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/policies")
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var entries []*ACLPolicyListEntry
if err := decodeBody(resp, &entries); err != nil {
return nil, nil, err
}
return entries, qm, nil
}
// RulesTranslate translates the legacy rule syntax into the current syntax.
//
// Deprecated: Support for the legacy syntax translation will be removed
// when legacy ACL support is removed.
func (a *ACL) RulesTranslate(rules io.Reader) (string, error) {
r := a.c.newRequest("POST", "/v1/acl/rules/translate")
r.body = 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
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return "", err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
ruleBytes, err := ioutil.ReadAll(resp.Body)
if err != nil {
return "", fmt.Errorf("Failed to read translated rule body: %v", err)
}
return string(ruleBytes), nil
}
// RulesTranslateToken translates the rules associated with the legacy syntax
// into the current syntax and returns the results.
//
// Deprecated: Support for the legacy syntax translation will be removed
// when legacy ACL support is removed.
func (a *ACL) RulesTranslateToken(tokenID string) (string, error) {
r := a.c.newRequest("GET", "/v1/acl/rules/translate/"+tokenID)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return "", err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
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
ruleBytes, err := ioutil.ReadAll(resp.Body)
if err != nil {
return "", fmt.Errorf("Failed to read translated rule body: %v", err)
}
return string(ruleBytes), 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
}
// RoleCreate will create a new role. It is not allowed for the role parameters
// ID field to be set as this will be generated by Consul while processing the request.
func (a *ACL) RoleCreate(role *ACLRole, q *WriteOptions) (*ACLRole, *WriteMeta, error) {
if role.ID != "" {
return nil, nil, fmt.Errorf("Cannot specify an ID in Role Creation")
}
r := a.c.newRequest("PUT", "/v1/acl/role")
r.setWriteOptions(q)
r.obj = role
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLRole
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// RoleUpdate updates a role. The ID field of the role parameter must be set to an
// existing role ID
func (a *ACL) RoleUpdate(role *ACLRole, q *WriteOptions) (*ACLRole, *WriteMeta, error) {
if role.ID == "" {
return nil, nil, fmt.Errorf("Must specify an ID in Role Update")
}
r := a.c.newRequest("PUT", "/v1/acl/role/"+role.ID)
r.setWriteOptions(q)
r.obj = role
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLRole
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// RoleDelete deletes a role given its ID.
func (a *ACL) RoleDelete(roleID string, q *WriteOptions) (*WriteMeta, error) {
r := a.c.newRequest("DELETE", "/v1/acl/role/"+roleID)
r.setWriteOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, err
}
resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
return wm, nil
}
// RoleRead retrieves the role details (by ID). Returns nil if not found.
func (a *ACL) RoleRead(roleID string, q *QueryOptions) (*ACLRole, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/role/"+roleID)
r.setQueryOptions(q)
found, rtt, resp, err := requireNotFoundOrOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
if !found {
return nil, qm, nil
}
var out ACLRole
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, qm, nil
}
// RoleReadByName retrieves the role details (by name). Returns nil if not found.
func (a *ACL) RoleReadByName(roleName string, q *QueryOptions) (*ACLRole, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/role/name/"+url.QueryEscape(roleName))
r.setQueryOptions(q)
found, rtt, resp, err := requireNotFoundOrOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
if !found {
return nil, qm, nil
}
var out ACLRole
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, qm, nil
}
// RoleList retrieves a listing of all roles. The listing does not include some
// metadata for the role as those should be retrieved by subsequent calls to
// RoleRead.
func (a *ACL) RoleList(q *QueryOptions) ([]*ACLRole, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/roles")
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var entries []*ACLRole
if err := decodeBody(resp, &entries); err != nil {
return nil, nil, err
}
return entries, qm, nil
}
// AuthMethodCreate will create a new auth method.
func (a *ACL) AuthMethodCreate(method *ACLAuthMethod, q *WriteOptions) (*ACLAuthMethod, *WriteMeta, error) {
if method.Name == "" {
return nil, nil, fmt.Errorf("Must specify a Name in Auth Method Creation")
}
r := a.c.newRequest("PUT", "/v1/acl/auth-method")
r.setWriteOptions(q)
r.obj = method
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLAuthMethod
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// AuthMethodUpdate updates an auth method.
func (a *ACL) AuthMethodUpdate(method *ACLAuthMethod, q *WriteOptions) (*ACLAuthMethod, *WriteMeta, error) {
if method.Name == "" {
return nil, nil, fmt.Errorf("Must specify a Name in Auth Method Update")
}
r := a.c.newRequest("PUT", "/v1/acl/auth-method/"+url.QueryEscape(method.Name))
r.setWriteOptions(q)
r.obj = method
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLAuthMethod
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// AuthMethodDelete deletes an auth method given its Name.
func (a *ACL) AuthMethodDelete(methodName string, q *WriteOptions) (*WriteMeta, error) {
if methodName == "" {
return nil, fmt.Errorf("Must specify a Name in Auth Method Delete")
}
r := a.c.newRequest("DELETE", "/v1/acl/auth-method/"+url.QueryEscape(methodName))
r.setWriteOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, err
}
resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
return wm, nil
}
// AuthMethodRead retrieves the auth method. Returns nil if not found.
func (a *ACL) AuthMethodRead(methodName string, q *QueryOptions) (*ACLAuthMethod, *QueryMeta, error) {
if methodName == "" {
return nil, nil, fmt.Errorf("Must specify a Name in Auth Method Read")
}
r := a.c.newRequest("GET", "/v1/acl/auth-method/"+url.QueryEscape(methodName))
r.setQueryOptions(q)
found, rtt, resp, err := requireNotFoundOrOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
if !found {
return nil, qm, nil
}
var out ACLAuthMethod
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, qm, nil
}
// AuthMethodList retrieves a listing of all auth methods. The listing does not
// include some metadata for the auth method as those should be retrieved by
// subsequent calls to AuthMethodRead.
func (a *ACL) AuthMethodList(q *QueryOptions) ([]*ACLAuthMethodListEntry, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/auth-methods")
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var entries []*ACLAuthMethodListEntry
if err := decodeBody(resp, &entries); err != nil {
return nil, nil, err
}
return entries, qm, nil
}
// BindingRuleCreate will create a new binding rule. It is not allowed for the
// binding rule parameter's ID field to be set as this will be generated by
// Consul while processing the request.
func (a *ACL) BindingRuleCreate(rule *ACLBindingRule, q *WriteOptions) (*ACLBindingRule, *WriteMeta, error) {
if rule.ID != "" {
return nil, nil, fmt.Errorf("Cannot specify an ID in Binding Rule Creation")
}
r := a.c.newRequest("PUT", "/v1/acl/binding-rule")
r.setWriteOptions(q)
r.obj = rule
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLBindingRule
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// BindingRuleUpdate updates a binding rule. The ID field of the role binding
// rule parameter must be set to an existing binding rule ID.
func (a *ACL) BindingRuleUpdate(rule *ACLBindingRule, q *WriteOptions) (*ACLBindingRule, *WriteMeta, error) {
if rule.ID == "" {
return nil, nil, fmt.Errorf("Must specify an ID in Binding Rule Update")
}
r := a.c.newRequest("PUT", "/v1/acl/binding-rule/"+rule.ID)
r.setWriteOptions(q)
r.obj = rule
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLBindingRule
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// BindingRuleDelete deletes a binding rule given its ID.
func (a *ACL) BindingRuleDelete(bindingRuleID string, q *WriteOptions) (*WriteMeta, error) {
r := a.c.newRequest("DELETE", "/v1/acl/binding-rule/"+bindingRuleID)
r.setWriteOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, err
}
resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
return wm, nil
}
// BindingRuleRead retrieves the binding rule details. Returns nil if not found.
func (a *ACL) BindingRuleRead(bindingRuleID string, q *QueryOptions) (*ACLBindingRule, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/binding-rule/"+bindingRuleID)
r.setQueryOptions(q)
found, rtt, resp, err := requireNotFoundOrOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
if !found {
return nil, qm, nil
}
var out ACLBindingRule
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, qm, nil
}
// BindingRuleList retrieves a listing of all binding rules.
func (a *ACL) BindingRuleList(methodName string, q *QueryOptions) ([]*ACLBindingRule, *QueryMeta, error) {
r := a.c.newRequest("GET", "/v1/acl/binding-rules")
if methodName != "" {
r.params.Set("authmethod", methodName)
}
r.setQueryOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
qm := &QueryMeta{}
parseQueryMeta(resp, qm)
qm.RequestTime = rtt
var entries []*ACLBindingRule
if err := decodeBody(resp, &entries); err != nil {
return nil, nil, err
}
return entries, qm, nil
}
// Login is used to exchange auth method credentials for a newly-minted Consul Token.
func (a *ACL) Login(auth *ACLLoginParams, q *WriteOptions) (*ACLToken, *WriteMeta, error) {
r := a.c.newRequest("POST", "/v1/acl/login")
r.setWriteOptions(q)
r.obj = auth
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, nil, err
}
defer resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
var out ACLToken
if err := decodeBody(resp, &out); err != nil {
return nil, nil, err
}
return &out, wm, nil
}
// Logout is used to destroy a Consul Token created via Login().
func (a *ACL) Logout(q *WriteOptions) (*WriteMeta, error) {
r := a.c.newRequest("POST", "/v1/acl/logout")
r.setWriteOptions(q)
rtt, resp, err := requireOK(a.c.doRequest(r))
if err != nil {
return nil, err
}
resp.Body.Close()
wm := &WriteMeta{RequestTime: rtt}
return wm, nil
}