Support rejoin after leave. Fixes #110.

CHANGELOG.md

@@ -12,6 +12,7 @@ IMPROVEMENTS:
   * Enable logging to syslog. [GH-105]
   * Allow raw key value lookup [GH-150]
   * Log encryption enabled [GH-151]
+  * Support `-rejoin` to rejoin a cluster after a previous leave. [GH-110]
 
 ## 0.2.1 (May 20, 2014)
 

command/agent/command.go

@@ -64,6 +64,8 @@ func (c *Command) readConfig() *Config {
 
 	cmdFlags.BoolVar(&cmdConfig.EnableSyslog, "syslog", false,
 		"enable logging to syslog facility")
+	cmdFlags.BoolVar(&cmdConfig.RejoinAfterLeave, "rejoin", false,
+		"enable re-joining after a previous leave")
 	cmdFlags.Var((*AppendSliceValue)(&cmdConfig.StartJoin), "join",
 		"address of agent to join on startup")
 
@@ -507,6 +509,7 @@ Options:
   -log-level=info          Log level of the agent.
   -node=hostname           Name of this node. Must be unique in the cluster
   -protocol=N              Sets the protocol version. Defaults to latest.
+  -rejoin                  Ignores a previous leave and attempts to rejoin the cluster.
   -server                  Switches agent to server mode.
   -syslog                  Enables logging to syslog
   -ui-dir=path             Path to directory containing the Web UI resources

command/agent/config.go

@@ -135,6 +135,12 @@ type Config struct {
 	// on linux and OSX. Other platforms will generate an error.
 	EnableSyslog bool `mapstructure:"enable_syslog"`
 
+	// RejoinAfterLeave controls our interaction with the cluster after leave.
+	// When set to false (default), a leave causes Consul to not rejoin
+	// the cluster until an explicit join is received. If this is set to
+	// true, we ignore the leave, and rejoin the cluster on start.
+	RejoinAfterLeave bool `mapstructure:"rejoin_after_leave"`
+
 	// AEInterval controls the anti-entropy interval. This is how often
 	// the agent attempts to reconcile it's local state with the server'
 	// representation of our state. Defaults to every 60s.
@@ -436,6 +442,9 @@ func MergeConfig(a, b *Config) *Config {
 	if b.EnableSyslog {
 		result.EnableSyslog = true
 	}
+	if b.RejoinAfterLeave {
+		result.RejoinAfterLeave = true
+	}
 
 	// Copy the start join addresses
 	result.StartJoin = make([]string, 0, len(a.StartJoin)+len(b.StartJoin))

command/agent/config_test.go

@@ -279,6 +279,17 @@ func TestDecodeConfig(t *testing.T) {
 	if !config.EnableSyslog {
 		t.Fatalf("bad: %#v", config)
 	}
+
+	// Rejoin
+	input = `{"rejoin_after_leave": true}`
+	config, err = DecodeConfig(bytes.NewReader([]byte(input)))
+	if err != nil {
+		t.Fatalf("err: %s", err)
+	}
+
+	if !config.RejoinAfterLeave {
+		t.Fatalf("bad: %#v", config)
+	}
 }
 
 func TestDecodeConfig_Service(t *testing.T) {
@@ -412,6 +423,7 @@ func TestMergeConfig(t *testing.T) {
 		StartJoin:        []string{"1.1.1.1"},
 		UiDir:            "/opt/consul-ui",
 		EnableSyslog:     true,
+		RejoinAfterLeave: true,
 	}
 
 	c := MergeConfig(a, b)

consul/client.go

@@ -127,6 +127,7 @@ func (c *Client) setupSerf(conf *serf.Config, ch chan serf.Event, path string) (
 	conf.EventCh = ch
 	conf.SnapshotPath = filepath.Join(c.config.DataDir, path)
 	conf.ProtocolVersion = protocolVersionMap[c.config.ProtocolVersion]
+	conf.RejoinAfterLeave = c.config.RejoinAfterLeave
 	if err := ensurePath(conf.SnapshotPath, false); err != nil {
 		return nil, err
 	}

consul/config.go

@@ -108,6 +108,12 @@ type Config struct {
 	// Must be provided to serve TLS connections.
 	KeyFile string
 
+	// RejoinAfterLeave controls our interaction with Serf.
+	// When set to false (default), a leave causes a Consul to not rejoin
+	// the cluster until an explicit join is received. If this is set to
+	// true, we ignore the leave, and rejoin the cluster on start.
+	RejoinAfterLeave bool
+
 	// ServerUp callback can be used to trigger a notification that
 	// a Consul server is now up and known about.
 	ServerUp func()

consul/server.go

@@ -220,6 +220,7 @@ func (s *Server) setupSerf(conf *serf.Config, ch chan serf.Event, path string, w
 	conf.EventCh = ch
 	conf.SnapshotPath = filepath.Join(s.config.DataDir, path)
 	conf.ProtocolVersion = protocolVersionMap[s.config.ProtocolVersion]
+	conf.RejoinAfterLeave = s.config.RejoinAfterLeave
 
 	// Until Consul supports this fully, we disable automatic resolution.
 	// When enabled, the Serf gossip may just turn off if we are the minority

website/source/docs/agent/options.html.markdown

@@ -92,6 +92,11 @@ The options below are all specified on the command-line.
   version. This should be set only when [upgrading](/docs/upgrading.html).
   You can view the protocol versions supported by Consul by running `consul -v`.
 
+* `-rejoin` - When provided Consul will ignore a previous leave and attempt to
+  rejoin the cluster when starting. By default, Consul treats leave as a permanent
+  intent, and does not attempt to join the cluster again when starting. This flag
+  allows the previous state to be used to rejoin the cluster.
+
 * `-server` - This flag is used to control if an agent is in server or client mode. When provided,
   an agent will act as a Consul server. Each Consul cluster must have at least one server, and ideally
   no more than 5 *per* datacenter. All servers participate in the Raft consensus algorithm, to ensure that
@@ -164,6 +169,8 @@ definitions support being updated during a reload.
 
 * `enable_syslog` - Equivalent to the `-syslog` command-line flag.
 
+* `rejoin_after_leave` - Equivalent to the `-rejoin` command-line flag.
+
 * `ca_file` - This provides a the file path to a PEM encoded certificate authority.
   The certificate authority is used to check the authenticity of client and server
   connections with the appropriate `verify_incoming` or `verify_outgoing` flags.