Add replication metrics (#10073)

# Conflicts:
#	agent/consul/replication.go
This commit is contained in:
Matt Keeler 2021-04-22 11:20:53 -04:00 committed by Matt Keeler
parent 40eeebd654
commit 835d3d4258
4 changed files with 43 additions and 4 deletions

View File

@ -790,10 +790,19 @@ func (s *Server) runLegacyACLReplication(ctx context.Context) error {
}
if err != nil {
metrics.SetGauge([]string{"leader", "replication", "acl-legacy", "status"},
0,
)
lastRemoteIndex = 0
s.updateACLReplicationStatusError()
legacyACLLogger.Warn("Legacy ACL replication error (will retry if still leader)", "error", err)
} else {
metrics.SetGauge([]string{"leader", "replication", "acl-legacy", "status"},
1,
)
metrics.SetGauge([]string{"leader", "replication", "acl-legacy", "index"},
float32(index),
)
lastRemoteIndex = index
s.updateACLReplicationStatusIndex(structs.ACLReplicateLegacy, index)
legacyACLLogger.Debug("Legacy ACL replication completed through remote index", "index", index)
@ -851,7 +860,7 @@ type replicateFunc func(ctx context.Context, logger hclog.Logger, lastRemoteInde
func (s *Server) runACLPolicyReplicator(ctx context.Context) error {
policyLogger := s.aclReplicationLogger(structs.ACLReplicatePolicies.SingularNoun())
policyLogger.Info("started ACL Policy replication")
return s.runACLReplicator(ctx, policyLogger, structs.ACLReplicatePolicies, s.replicateACLPolicies)
return s.runACLReplicator(ctx, policyLogger, structs.ACLReplicatePolicies, s.replicateACLPolicies, "acl-policies")
}
// This function is only intended to be run as a managed go routine, it will block until
@ -859,7 +868,7 @@ func (s *Server) runACLPolicyReplicator(ctx context.Context) error {
func (s *Server) runACLRoleReplicator(ctx context.Context) error {
roleLogger := s.aclReplicationLogger(structs.ACLReplicateRoles.SingularNoun())
roleLogger.Info("started ACL Role replication")
return s.runACLReplicator(ctx, roleLogger, structs.ACLReplicateRoles, s.replicateACLRoles)
return s.runACLReplicator(ctx, roleLogger, structs.ACLReplicateRoles, s.replicateACLRoles, "acl-roles")
}
// This function is only intended to be run as a managed go routine, it will block until
@ -867,7 +876,7 @@ func (s *Server) runACLRoleReplicator(ctx context.Context) error {
func (s *Server) runACLTokenReplicator(ctx context.Context) error {
tokenLogger := s.aclReplicationLogger(structs.ACLReplicateTokens.SingularNoun())
tokenLogger.Info("started ACL Token replication")
return s.runACLReplicator(ctx, tokenLogger, structs.ACLReplicateTokens, s.replicateACLTokens)
return s.runACLReplicator(ctx, tokenLogger, structs.ACLReplicateTokens, s.replicateACLTokens, "acl-tokens")
}
// This function is only intended to be run as a managed go routine, it will block until
@ -877,6 +886,7 @@ func (s *Server) runACLReplicator(
logger hclog.Logger,
replicationType structs.ACLReplicationType,
replicateFunc replicateFunc,
metricName string,
) error {
var failedAttempts uint
limiter := rate.NewLimiter(rate.Limit(s.config.ACLReplicationRate), s.config.ACLReplicationBurst)
@ -897,6 +907,9 @@ func (s *Server) runACLReplicator(
}
if err != nil {
metrics.SetGauge([]string{"leader", "replication", metricName, "status"},
0,
)
lastRemoteIndex = 0
s.updateACLReplicationStatusError()
logger.Warn("ACL replication error (will retry if still leader)",
@ -913,6 +926,12 @@ func (s *Server) runACLReplicator(
// do nothing
}
} else {
metrics.SetGauge([]string{"leader", "replication", metricName, "status"},
1,
)
metrics.SetGauge([]string{"leader", "replication", metricName, "index"},
float32(index),
)
lastRemoteIndex = index
s.updateACLReplicationStatusIndex(replicationType, index)
logger.Debug("ACL replication completed through remote index",

View File

@ -23,6 +23,7 @@ const (
type ReplicatorDelegate interface {
Replicate(ctx context.Context, lastRemoteIndex uint64, logger hclog.Logger) (index uint64, exit bool, err error)
MetricName() string
}
type ReplicatorConfig struct {
@ -104,6 +105,9 @@ func (r *Replicator) Run(ctx context.Context) error {
}
if err != nil {
metrics.SetGauge([]string{"leader", "replication", r.delegate.MetricName(), "status"},
0,
)
// reset the lastRemoteIndex when there is an RPC failure. This should cause a full sync to be done during
// the next round of replication
atomic.StoreUint64(&r.lastRemoteIndex, 0)
@ -112,6 +116,12 @@ func (r *Replicator) Run(ctx context.Context) error {
r.logger.Warn("replication error (will retry if still leader)", "error", err)
}
} else {
metrics.SetGauge([]string{"leader", "replication", r.delegate.MetricName(), "status"},
1,
)
metrics.SetGauge([]string{"leader", "replication", r.delegate.MetricName(), "index"},
float32(index),
)
atomic.StoreUint64(&r.lastRemoteIndex, index)
r.logger.Debug("replication completed through remote index", "index", index)
}
@ -134,6 +144,11 @@ type ReplicatorFunc func(ctx context.Context, lastRemoteIndex uint64, logger hcl
type FunctionReplicator struct {
ReplicateFn ReplicatorFunc
Name string
}
func (r *FunctionReplicator) MetricName() string {
return r.Name
}
func (r *FunctionReplicator) Replicate(ctx context.Context, lastRemoteIndex uint64, logger hclog.Logger) (uint64, bool, error) {
@ -177,6 +192,10 @@ type IndexReplicator struct {
Logger hclog.Logger
}
func (r *IndexReplicator) MetricName() string {
return r.Delegate.MetricName()
}
func (r *IndexReplicator) Replicate(ctx context.Context, lastRemoteIndex uint64, _ hclog.Logger) (uint64, bool, error) {
fetchStart := time.Now()
lenRemote, remote, remoteIndex, err := r.Delegate.FetchRemote(lastRemoteIndex)

View File

@ -20,6 +20,7 @@ func TestReplicationRestart(t *testing.T) {
ReplicateFn: func(ctx context.Context, lastRemoteIndex uint64, logger hclog.Logger) (uint64, bool, error) {
return 1, false, nil
},
Name: "foo",
},
Rate: 1,

View File

@ -412,7 +412,7 @@ func NewServer(config *Config, options ...ConsulOption) (*Server, error) {
configReplicatorConfig := ReplicatorConfig{
Name: logging.ConfigEntry,
Delegate: &FunctionReplicator{ReplicateFn: s.replicateConfig},
Delegate: &FunctionReplicator{ReplicateFn: s.replicateConfig, Name: "config-entries"},
Rate: s.config.ConfigReplicationRate,
Burst: s.config.ConfigReplicationBurst,
Logger: s.logger,