consul: Adding telemetry

consul/catalog_endpoint.go

@@ -2,7 +2,9 @@ package consul
 
 import (
 	"fmt"
+	"github.com/armon/go-metrics"
 	"github.com/hashicorp/consul/consul/structs"
+	"time"
 )
 
 // Catalog endpoint is used to manipulate the service catalog
@@ -15,6 +17,7 @@ func (c *Catalog) Register(args *structs.RegisterRequest, reply *struct{}) error
 	if done, err := c.srv.forward("Catalog.Register", args.Datacenter, args, reply); done {
 		return err
 	}
+	defer metrics.MeasureSince([]string{"consul", "catalog", "register"}, time.Now())
 
 	// Verify the args
 	if args.Node == "" || args.Address == "" {
@@ -55,6 +58,7 @@ func (c *Catalog) Deregister(args *structs.DeregisterRequest, reply *struct{}) e
 	if done, err := c.srv.forward("Catalog.Deregister", args.Datacenter, args, reply); done {
 		return err
 	}
+	defer metrics.MeasureSince([]string{"consul", "catalog", "deregister"}, time.Now())
 
 	// Verify the args
 	if args.Node == "" {

consul/leader.go

@@ -1,6 +1,7 @@
 package consul
 
 import (
+	"github.com/armon/go-metrics"
 	"github.com/hashicorp/consul/consul/structs"
 	"github.com/hashicorp/raft"
 	"github.com/hashicorp/serf/serf"
@@ -59,11 +60,13 @@ RECONCILE:
 	interval := time.After(s.config.ReconcileInterval)
 
 	// Apply a raft barrier to ensure our FSM is caught up
+	start := time.Now()
 	barrier := s.raft.Barrier(0)
 	if err := barrier.Error(); err != nil {
 		s.logger.Printf("[ERR] consul: failed to wait for barrier: %v", err)
 		goto WAIT
 	}
+	metrics.MeasureSince([]string{"consul", "leader", "barrier"}, start)
 
 	// Reconcile any missing data
 	if err := s.reconcile(); err != nil {
@@ -97,6 +100,7 @@ WAIT:
 // Mainly we need to ensure all live nodes are registered, all failed
 // nodes are marked as such, and all left nodes are de-registered.
 func (s *Server) reconcile() (err error) {
+	defer metrics.MeasureSince([]string{"consul", "leader", "reconcile"}, time.Now())
 	members := s.serfLAN.Members()
 	for _, member := range members {
 		if err := s.reconcileMember(member); err != nil {
@@ -114,6 +118,7 @@ func (s *Server) reconcileMember(member serf.Member) error {
 		s.logger.Printf("[WARN] consul: skipping reconcile of node %v", member)
 		return nil
 	}
+	defer metrics.MeasureSince([]string{"consul", "leader", "reconcileMember"}, time.Now())
 	var err error
 	switch member.Status {
 	case serf.StatusAlive:

consul/rpc.go

@@ -2,6 +2,7 @@ package consul
 
 import (
 	"fmt"
+	"github.com/armon/go-metrics"
 	"github.com/hashicorp/consul/consul/structs"
 	"github.com/inconshreveable/muxado"
 	"github.com/ugorji/go/codec"
@@ -43,6 +44,7 @@ func (s *Server) listen() {
 		s.rpcClientLock.Unlock()
 
 		go s.handleConn(conn)
+		metrics.IncrCounter([]string{"consul", "rpc", "accept_conn"}, 1)
 	}
 }
 
@@ -63,6 +65,7 @@ func (s *Server) handleConn(conn net.Conn) {
 		s.handleConsulConn(conn)
 
 	case rpcRaft:
+		metrics.IncrCounter([]string{"consul", "rpc", "raft_handoff"}, 1)
 		s.raftLayer.Handoff(conn)
 
 	case rpcMultiplex:
@@ -154,6 +157,7 @@ func (s *Server) forwardDC(method, dc string, args interface{}, reply interface{
 	s.remoteLock.RUnlock()
 
 	// Forward to remote Consul
+	metrics.IncrCounter([]string{"consul", "rpc", "cross-dc", dc}, 1)
 	return s.connPool.RPC(server, method, args, reply)
 }
 

consul/server.go

@@ -17,7 +17,6 @@ const (
 	serfLANSnapshot = "serf/local.snapshot"
 	serfWANSnapshot = "serf/remote.snapshot"
 	raftState       = "raft/"
-	bootstrapFlag   = "b"
 )
 
 // Server is Consul server which manages the service discovery,