From 7d3640329158e857b32cfedea7d9f84897b6e074 Mon Sep 17 00:00:00 2001
From: Kyle Havlovitz <kylehav@gmail.com>
Date: Wed, 12 Apr 2017 15:28:18 -0700
Subject: [PATCH] Wait to initialize autopilot until all servers are >= 0.8.0

---
 consul/autopilot.go      | 29 ++++++++++++++++++++++++++++-
 consul/autopilot_test.go |  2 ++
 consul/leader.go         | 11 ++++++++---
 3 files changed, 38 insertions(+), 4 deletions(-)

diff --git a/consul/autopilot.go b/consul/autopilot.go
index 7ea138566e..0a3294bf49 100644
--- a/consul/autopilot.go
+++ b/consul/autopilot.go
@@ -10,6 +10,7 @@ import (
 	"github.com/armon/go-metrics"
 	"github.com/hashicorp/consul/consul/agent"
 	"github.com/hashicorp/consul/consul/structs"
+	"github.com/hashicorp/go-version"
 	"github.com/hashicorp/raft"
 	"github.com/hashicorp/serf/serf"
 )
@@ -33,6 +34,8 @@ func (s *Server) stopAutopilot() {
 	s.autopilotWaitGroup.Wait()
 }
 
+var minAutopilotVersion, _ = version.NewVersion("0.8.0")
+
 // autopilotLoop periodically looks for nonvoting servers to promote and dead servers to remove.
 func (s *Server) autopilotLoop() {
 	defer s.autopilotWaitGroup.Done()
@@ -53,6 +56,15 @@ func (s *Server) autopilotLoop() {
 				break
 			}
 
+			// Setup autopilot config if we need to
+			if autopilotConf == nil {
+				if err := s.initializeAutopilot(); err != nil {
+					s.logger.Printf("[ERR] autopilot: %v", err)
+				}
+
+				continue
+			}
+
 			if err := s.autopilotPolicy.PromoteNonVoters(autopilotConf); err != nil {
 				s.logger.Printf("[ERR] autopilot: error checking for non-voters to promote: %s", err)
 			}
@@ -68,11 +80,26 @@ func (s *Server) autopilotLoop() {
 	}
 }
 
+// lowestServerVersion returns the lowest version among the alive servers
+func (s *Server) lowestServerVersion() *version.Version {
+	lowest := minAutopilotVersion
+
+	for _, member := range s.LANMembers() {
+		if valid, parts := agent.IsConsulServer(member); valid && parts.Status == serf.StatusAlive {
+			if parts.Build.LessThan(lowest) {
+				lowest = &parts.Build
+			}
+		}
+	}
+
+	return lowest
+}
+
 // pruneDeadServers removes up to numPeers/2 failed servers
 func (s *Server) pruneDeadServers() error {
 	state := s.fsm.State()
 	_, autopilotConf, err := state.AutopilotConfig()
-	if err != nil {
+	if err != nil || autopilotConf == nil {
 		return err
 	}
 
diff --git a/consul/autopilot_test.go b/consul/autopilot_test.go
index 5aed5719ba..744e05f7f5 100644
--- a/consul/autopilot_test.go
+++ b/consul/autopilot_test.go
@@ -191,6 +191,8 @@ func TestAutopilot_CleanupStaleRaftServer(t *testing.T) {
 		}
 	}
 
+	testutil.WaitForLeader(t, s1.RPC, "dc1")
+
 	// Add s4 to peers directly
 	s4addr := fmt.Sprintf("127.0.0.1:%d",
 		s4.config.SerfLANConfig.MemberlistConfig.BindPort)
diff --git a/consul/leader.go b/consul/leader.go
index dc41a48792..bf48048e41 100644
--- a/consul/leader.go
+++ b/consul/leader.go
@@ -153,10 +153,9 @@ func (s *Server) establishLeadership() error {
 		return err
 	}
 
-	// Setup autopilot config if we are the leader and need to
+	// Setup autopilot config if we need to
 	if err := s.initializeAutopilot(); err != nil {
-		s.logger.Printf("[ERR] consul: Autopilot initialization failed: %v", err)
-		return err
+		s.logger.Printf("[ERR] autopilot: %v", err)
 	}
 
 	s.startAutopilot()
@@ -252,6 +251,12 @@ func (s *Server) initializeACL() error {
 // initializeAutopilot is used to setup the autopilot config if we are
 // the leader and need to do this
 func (s *Server) initializeAutopilot() error {
+	lowestVersion := s.lowestServerVersion()
+
+	if !lowestVersion.Equal(minAutopilotVersion) && !lowestVersion.GreaterThan(minAutopilotVersion) {
+		return fmt.Errorf("can't initialize autopilot until all servers are >= %s", minAutopilotVersion.String())
+	}
+
 	// Bail if the config has already been initialized
 	state := s.fsm.State()
 	_, config, err := state.AutopilotConfig()