Reworked RAFT's leader change channel

Signed-off-by: Ivan Kozlovic <ivan@synadia.com>

server/raft.go

@@ -194,8 +194,8 @@ type raft struct {
 	apply    *ipQueue // of *CommittedEntry
 	reqs     *ipQueue // of *voteRequest
 	votes    *ipQueue // of *voteResponse
-	leadc    chan bool
 	stepdown *ipQueue // of string
+	leadc    chan bool
 	quit     chan struct{}
 }
 
@@ -378,8 +378,8 @@ func (s *Server) startRaftNode(cfg *RaftConfig) (RaftNode, error) {
 		entry:    newIPQueue(), // of *appendEntry
 		resp:     newIPQueue(), // of *appendEntryResponse
 		apply:    newIPQueue(), // of *CommittedEntry
-		leadc:    make(chan bool, 8),
 		stepdown: newIPQueue(), // of string
+		leadc:    make(chan bool, 1),
 		observer: cfg.Observer,
 		extSt:    ps.domainExt,
 	}
@@ -3285,14 +3285,19 @@ func (n *raft) quorumNeeded() int {
 
 // Lock should be held.
 func (n *raft) updateLeadChange(isLeader bool) {
-	select {
-	case n.leadc <- isLeader:
-	case <-n.leadc:
-		// We had an old value not consumed.
+	// We don't care about values that have not been consumed (transitory states),
+	// so we dequeue any state that is pending and push the new one.
+	for {
 		select {
 		case n.leadc <- isLeader:
+			return
 		default:
-			n.error("Failed to post lead change to %v for %q", isLeader, n.group)
+			select {
+			case <-n.leadc:
+			default:
+				// May have been consumed by the "reader" go routine, so go back
+				// to the top of the loop and try to send again.
+			}
 		}
 	}
 }