Merge pull request #1276 from hashicorp/f-consul-server-autojoin

Teach Nomad servers how to fall back to Consul.

client/client.go

@@ -39,7 +39,7 @@ const (
 
 	// datacenterQueryLimit searches through up to this many adjacent
 	// datacenters looking for the Nomad server service.
-	datacenterQueryLimit = 5
+	datacenterQueryLimit = 9
 
 	// registerRetryIntv is minimum interval on which we retry
 	// registration. We pick a value between this and 2x this.
@@ -1253,8 +1253,8 @@ func (c *Client) setupConsulSyncer() error {
 			// a new set of servers so it's okay.
 			nearestDC := dcs[0]
 			otherDCs := make([]string, 0, len(dcs))
-			otherDCs = dcs[1:lib.MinInt(len(dcs), datacenterQueryLimit)]
 			shuffleStrings(otherDCs)
+			otherDCs = dcs[1:lib.MinInt(len(dcs), datacenterQueryLimit)]
 
 			dcs = append([]string{nearestDC}, otherDCs...)
 		}
@@ -1270,7 +1270,7 @@ func (c *Client) setupConsulSyncer() error {
 		var mErr multierror.Error
 		const defaultMaxNumNomadServers = 8
 		nomadServerServices := make([]string, 0, defaultMaxNumNomadServers)
-		c.logger.Printf("[DEBUG] client.consul: bootstrap contacting following Consul DCs: %q", dcs)
+		c.logger.Printf("[DEBUG] client.consul: bootstrap contacting following Consul DCs: %+q", dcs)
 		for _, dc := range dcs {
 			consulOpts := &consulapi.QueryOptions{
 				AllowStale: true,

command/agent/agent.go

@@ -11,6 +11,7 @@ import (
 	"strconv"
 	"strings"
 	"sync"
+	"sync/atomic"
 	"time"
 
 	"github.com/hashicorp/nomad/client"
@@ -110,7 +111,7 @@ func (a *Agent) serverConfig() (*nomad.Config, error) {
 		if a.config.Server.BootstrapExpect == 1 {
 			conf.Bootstrap = true
 		} else {
-			conf.BootstrapExpect = a.config.Server.BootstrapExpect
+			atomic.StoreInt32(&conf.BootstrapExpect, int32(a.config.Server.BootstrapExpect))
 		}
 	}
 	if a.config.DataDir != "" {
@@ -235,7 +236,7 @@ func (a *Agent) serverConfig() (*nomad.Config, error) {
 		return nil, fmt.Errorf("server_service_name must be set when auto_advertise is enabled")
 	}
 
-	// conf.ConsulConfig = a.config.Consul
+	conf.ConsulConfig = a.config.Consul
 
 	return conf, nil
 }
@@ -377,7 +378,7 @@ func (a *Agent) setupServer() error {
 	}
 
 	// Create the server
-	server, err := nomad.NewServer(conf)
+	server, err := nomad.NewServer(conf, a.consulSyncer)
 	if err != nil {
 		return fmt.Errorf("server setup failed: %v", err)
 	}

nomad/config.go

@@ -10,6 +10,7 @@ import (
 
 	"github.com/hashicorp/memberlist"
 	"github.com/hashicorp/nomad/nomad/structs"
+	"github.com/hashicorp/nomad/nomad/structs/config"
 	"github.com/hashicorp/nomad/scheduler"
 	"github.com/hashicorp/raft"
 	"github.com/hashicorp/serf/serf"
@@ -51,8 +52,9 @@ type Config struct {
 
 	// BootstrapExpect mode is used to automatically bring up a
 	// collection of Nomad servers. This can be used to automatically
-	// bring up a collection of nodes.
-	BootstrapExpect int
+	// bring up a collection of nodes.  All operations on BootstrapExpect
+	// must be handled via `atomic.*Int32()` calls.
+	BootstrapExpect int32
 
 	// DataDir is the directory to store our state in
 	DataDir string
@@ -176,6 +178,9 @@ type Config struct {
 	// a new leader is elected, since we no longer know the status
 	// of all the heartbeats.
 	FailoverHeartbeatTTL time.Duration
+
+	// ConsulConfig is this Agent's Consul configuration
+	ConsulConfig *config.ConsulConfig
 }
 
 // CheckVersion is used to check if the ProtocolVersion is valid

nomad/serf.go

@@ -1,6 +1,10 @@
 package nomad
 
-import "github.com/hashicorp/serf/serf"
+import (
+	"sync/atomic"
+
+	"github.com/hashicorp/serf/serf"
+)
 
 const (
 	// StatusReap is used to update the status of a node if we
@@ -66,7 +70,7 @@ func (s *Server) nodeJoin(me serf.MemberEvent) {
 		s.peerLock.Unlock()
 
 		// If we still expecting to bootstrap, may need to handle this
-		if s.config.BootstrapExpect != 0 {
+		if atomic.LoadInt32(&s.config.BootstrapExpect) != 0 {
 			s.maybeBootstrap()
 		}
 	}
@@ -91,7 +95,7 @@ func (s *Server) maybeBootstrap() {
 	// Bootstrap can only be done if there are no committed logs,
 	// remove our expectations of bootstrapping
 	if index != 0 {
-		s.config.BootstrapExpect = 0
+		atomic.StoreInt32(&s.config.BootstrapExpect, 0)
 		return
 	}
 
@@ -106,7 +110,7 @@ func (s *Server) maybeBootstrap() {
 		if p.Region != s.config.Region {
 			continue
 		}
-		if p.Expect != 0 && p.Expect != s.config.BootstrapExpect {
+		if p.Expect != 0 && p.Expect != int(atomic.LoadInt32(&s.config.BootstrapExpect)) {
 			s.logger.Printf("[ERR] nomad: peer %v has a conflicting expect value. All nodes should expect the same number.", member)
 			return
 		}
@@ -118,7 +122,7 @@ func (s *Server) maybeBootstrap() {
 	}
 
 	// Skip if we haven't met the minimum expect count
-	if len(addrs) < s.config.BootstrapExpect {
+	if len(addrs) < int(atomic.LoadInt32(&s.config.BootstrapExpect)) {
 		return
 	}
 
@@ -128,8 +132,8 @@ func (s *Server) maybeBootstrap() {
 		s.logger.Printf("[ERR] nomad: failed to bootstrap peers: %v", err)
 	}
 
-	// Bootstrapping comlete, don't enter this again
-	s.config.BootstrapExpect = 0
+	// Bootstrapping complete, don't enter this again
+	atomic.StoreInt32(&s.config.BootstrapExpect, 0)
 }
 
 // nodeFailed is used to handle fail events on the serf cluster