open-consul/agent/consul/txn_endpoint.go

223 lines
6.2 KiB
Go
Raw Normal View History

package consul
import (
"fmt"
"time"
"github.com/armon/go-metrics"
"github.com/armon/go-metrics/prometheus"
"github.com/hashicorp/go-hclog"
"github.com/hashicorp/consul/acl"
"github.com/hashicorp/consul/agent/structs"
"github.com/hashicorp/consul/api"
)
var TxnSummaries = []prometheus.SummaryDefinition{
{
Name: []string{"txn", "apply"},
2020-11-16 19:02:11 +00:00
Help: "Measures the time spent applying a transaction operation.",
},
{
Name: []string{"txn", "read"},
2020-11-16 19:02:11 +00:00
Help: "Measures the time spent returning a read transaction.",
},
}
// Txn endpoint is used to perform multi-object atomic transactions.
type Txn struct {
srv *Server
logger hclog.Logger
}
// preCheck is used to verify the incoming operations before any further
// processing takes place. This checks things like ACLs.
New ACLs (#4791) This PR is almost a complete rewrite of the ACL system within Consul. It brings the features more in line with other HashiCorp products. Obviously there is quite a bit left to do here but most of it is related docs, testing and finishing the last few commands in the CLI. I will update the PR description and check off the todos as I finish them over the next few days/week. Description At a high level this PR is mainly to split ACL tokens from Policies and to split the concepts of Authorization from Identities. A lot of this PR is mostly just to support CRUD operations on ACLTokens and ACLPolicies. These in and of themselves are not particularly interesting. The bigger conceptual changes are in how tokens get resolved, how backwards compatibility is handled and the separation of policy from identity which could lead the way to allowing for alternative identity providers. On the surface and with a new cluster the ACL system will look very similar to that of Nomads. Both have tokens and policies. Both have local tokens. The ACL management APIs for both are very similar. I even ripped off Nomad's ACL bootstrap resetting procedure. There are a few key differences though. Nomad requires token and policy replication where Consul only requires policy replication with token replication being opt-in. In Consul local tokens only work with token replication being enabled though. All policies in Nomad are globally applicable. In Consul all policies are stored and replicated globally but can be scoped to a subset of the datacenters. This allows for more granular access management. Unlike Nomad, Consul has legacy baggage in the form of the original ACL system. The ramifications of this are: A server running the new system must still support other clients using the legacy system. A client running the new system must be able to use the legacy RPCs when the servers in its datacenter are running the legacy system. The primary ACL DC's servers running in legacy mode needs to be a gate that keeps everything else in the entire multi-DC cluster running in legacy mode. So not only does this PR implement the new ACL system but has a legacy mode built in for when the cluster isn't ready for new ACLs. Also detecting that new ACLs can be used is automatic and requires no configuration on the part of administrators. This process is detailed more in the "Transitioning from Legacy to New ACL Mode" section below.
2018-10-19 16:04:07 +00:00
func (t *Txn) preCheck(authorizer acl.Authorizer, ops structs.TxnOps) structs.TxnErrors {
var errors structs.TxnErrors
// Perform the pre-apply checks for any KV operations.
for i, op := range ops {
switch {
case op.KV != nil:
ok, err := kvsPreApply(t.logger, t.srv, authorizer, op.KV.Verb, &op.KV.DirEnt)
if err != nil {
2017-03-23 23:05:35 +00:00
errors = append(errors, &structs.TxnError{
OpIndex: i,
What: err.Error(),
})
} else if !ok {
err = fmt.Errorf("failed to lock key %q due to lock delay", op.KV.DirEnt.Key)
2017-03-23 23:05:35 +00:00
errors = append(errors, &structs.TxnError{
OpIndex: i,
What: err.Error(),
})
}
case op.Node != nil:
// Skip the pre-apply checks if this is a GET.
if op.Node.Verb == api.NodeGet {
break
}
node := op.Node.Node
if err := nodePreApply(node.Node, string(node.ID)); err != nil {
errors = append(errors, &structs.TxnError{
OpIndex: i,
What: err.Error(),
})
break
}
// Check that the token has permissions for the given operation.
if err := vetNodeTxnOp(op.Node, authorizer); err != nil {
errors = append(errors, &structs.TxnError{
OpIndex: i,
What: err.Error(),
})
}
case op.Service != nil:
// Skip the pre-apply checks if this is a GET.
if op.Service.Verb == api.ServiceGet {
break
}
service := &op.Service.Service
if err := servicePreApply(service, authorizer, op.Service.FillAuthzContext); err != nil {
errors = append(errors, &structs.TxnError{
OpIndex: i,
What: err.Error(),
})
}
case op.Check != nil:
// Skip the pre-apply checks if this is a GET.
if op.Check.Verb == api.CheckGet {
break
}
checkPreApply(&op.Check.Check)
// Check that the token has permissions for the given operation.
if err := vetCheckTxnOp(op.Check, authorizer); err != nil {
errors = append(errors, &structs.TxnError{
OpIndex: i,
What: err.Error(),
})
}
}
}
return errors
}
// vetNodeTxnOp applies the given ACL policy to a node transaction operation.
func vetNodeTxnOp(op *structs.TxnNodeOp, authz acl.Authorizer) error {
var authzContext acl.AuthorizerContext
op.FillAuthzContext(&authzContext)
if err := authz.ToAllowAuthorizer().NodeWriteAllowed(op.Node.Node, &authzContext); err != nil {
return err
}
return nil
}
// vetCheckTxnOp applies the given ACL policy to a check transaction operation.
func vetCheckTxnOp(op *structs.TxnCheckOp, authz acl.Authorizer) error {
var authzContext acl.AuthorizerContext
op.FillAuthzContext(&authzContext)
if op.Check.ServiceID == "" {
// Node-level check.
if err := authz.ToAllowAuthorizer().NodeWriteAllowed(op.Check.Node, &authzContext); err != nil {
return err
}
} else {
// Service-level check.
if err := authz.ToAllowAuthorizer().ServiceWriteAllowed(op.Check.ServiceName, &authzContext); err != nil {
return err
}
}
return nil
}
// Apply is used to apply multiple operations in a single, atomic transaction.
func (t *Txn) Apply(args *structs.TxnRequest, reply *structs.TxnResponse) error {
if done, err := t.srv.ForwardRPC("Txn.Apply", args, reply); done {
return err
}
defer metrics.MeasureSince([]string{"txn", "apply"}, time.Now())
// Run the pre-checks before we send the transaction into Raft.
authz, err := t.srv.ResolveToken(args.Token)
if err != nil {
return err
}
reply.Errors = t.preCheck(authz, args.Ops)
if len(reply.Errors) > 0 {
return nil
}
// Apply the update.
resp, err := t.srv.raftApply(structs.TxnRequestType, args)
if err != nil {
return fmt.Errorf("raft apply failed: %w", err)
}
// Convert the return type. This should be a cheap copy since we are
// just taking the two slices.
if txnResp, ok := resp.(structs.TxnResponse); ok {
txnResp.Results = FilterTxnResults(authz, txnResp.Results)
*reply = txnResp
} else {
return fmt.Errorf("unexpected return type %T", resp)
}
return nil
}
// Read is used to perform a read-only transaction that doesn't modify the state
// store. This is much more scalable since it doesn't go through Raft and
// supports staleness, so this should be preferred if you're just performing
// reads.
func (t *Txn) Read(args *structs.TxnReadRequest, reply *structs.TxnReadResponse) error {
if done, err := t.srv.ForwardRPC("Txn.Read", args, reply); done {
return err
}
defer metrics.MeasureSince([]string{"txn", "read"}, time.Now())
// We have to do this ourselves since we are not doing a blocking RPC.
if args.RequireConsistent {
if err := t.srv.consistentRead(); err != nil {
return err
}
}
// Run the pre-checks before we perform the read.
authz, err := t.srv.ResolveToken(args.Token)
if err != nil {
return err
}
// There are currently two different ways we handle permission issues.
//
// For simple reads such as KVGet and KVGetTree, the txn succeeds but the
// offending results are omitted. For more involved operations such as
// KVCheckIndex, the txn fails and permission denied errors are returned.
//
// TODO: Maybe we should unify these, or at least cover it in the docs?
reply.Errors = t.preCheck(authz, args.Ops)
if len(reply.Errors) > 0 {
return nil
}
// Run the read transaction.
state := t.srv.fsm.State()
reply.Results, reply.Errors = state.TxnRO(args.Ops)
total := len(reply.Results)
reply.Results = FilterTxnResults(authz, reply.Results)
reply.QueryMeta.ResultsFilteredByACLs = total != len(reply.Results)
// We have to do this ourselves since we are not doing a blocking RPC.
t.srv.setQueryMeta(&reply.QueryMeta, args.Token)
return nil
}