raft: send up-to-date commit index in heartbeats

TODO: describe why it is now safe

Signed-off-by: Pavel Kalinnikov <[email protected]>

raft.go

@@ -672,21 +672,12 @@ func (r *raft) maybeSendAppend(to uint64, sendIfEmpty bool) bool {
 
 // sendHeartbeat sends a heartbeat RPC to the given peer.
 func (r *raft) sendHeartbeat(to uint64, ctx []byte) {
-	// Attach the commit as min(to.matched, r.committed).
-	// When the leader sends out heartbeat message,
-	// the receiver(follower) might not be matched with the leader
-	// or it might not have all the committed entries.
-	// The leader MUST NOT forward the follower's commit to
-	// an unmatched index.
-	commit := min(r.trk.Progress[to].Match, r.raftLog.committed)
-	m := pb.Message{
+	r.send(pb.Message{
 		To:      to,
 		Type:    pb.MsgHeartbeat,
-		Commit:  commit,
+		Commit:  r.raftLog.committed,
 		Context: ctx,
-	}
-
-	r.send(m)
+	})
 }
 
 // bcastAppend sends RPC, with entries to all peers that are not up-to-date

testdata/async_storage_writes_append_aba_race.txt

@@ -346,16 +346,16 @@ process-ready 4
 ----
 Ready MustSync=false:
 Messages:
-4->1 MsgHeartbeat Term:3 Log:0/0
-4->2 MsgHeartbeat Term:3 Log:0/0
-4->3 MsgHeartbeat Term:3 Log:0/0
-4->5 MsgHeartbeat Term:3 Log:0/0
-4->6 MsgHeartbeat Term:3 Log:0/0
-4->7 MsgHeartbeat Term:3 Log:0/0
+4->1 MsgHeartbeat Term:3 Log:0/0 Commit:11
+4->2 MsgHeartbeat Term:3 Log:0/0 Commit:11
+4->3 MsgHeartbeat Term:3 Log:0/0 Commit:11
+4->5 MsgHeartbeat Term:3 Log:0/0 Commit:11
+4->6 MsgHeartbeat Term:3 Log:0/0 Commit:11
+4->7 MsgHeartbeat Term:3 Log:0/0 Commit:11
 
 deliver-msgs 1
 ----
-4->1 MsgHeartbeat Term:3 Log:0/0
+4->1 MsgHeartbeat Term:3 Log:0/0 Commit:11
 INFO 1 [term: 2] received a MsgHeartbeat message with higher term from 4 [term: 3]
 INFO 1 became follower at term 3
 

testdata/lagging_commit.txt

@@ -134,7 +134,7 @@ process-ready 1
 Ready MustSync=false:
 Messages:
 1->2 MsgHeartbeat Term:1 Log:0/0 Commit:13
-1->3 MsgHeartbeat Term:1 Log:0/0 Commit:11
+1->3 MsgHeartbeat Term:1 Log:0/0 Commit:13
 
 # Since the heartbeat message does not bump the follower's commit index, it will
 # take another roundtrip with the leader to update it. As such, the total time
@@ -146,12 +146,18 @@ Messages:
 # leader sent the up-to-date commit index in the heartbeat message.
 #
 # See https://github.com/etcd-io/raft/issues/138 which aims to fix this.
+#
+# Now this is fixed!
 stabilize 1 3
 ----
 > 3 receiving messages
-  1->3 MsgHeartbeat Term:1 Log:0/0 Commit:11
+  1->3 MsgHeartbeat Term:1 Log:0/0 Commit:13
 > 3 handling Ready
   Ready MustSync=false:
+  HardState Term:1 Vote:1 Commit:13
+  CommittedEntries:
+  1/12 EntryNormal "data1"
+  1/13 EntryNormal "data2"
   Messages:
   3->1 MsgHeartbeatResp Term:1 Log:0/0
 > 1 receiving messages
@@ -164,10 +170,6 @@ stabilize 1 3
   1->3 MsgApp Term:1 Log:1/13 Commit:13
 > 3 handling Ready
   Ready MustSync=false:
-  HardState Term:1 Vote:1 Commit:13
-  CommittedEntries:
-  1/12 EntryNormal "data1"
-  1/13 EntryNormal "data2"
   Messages:
   3->1 MsgAppResp Term:1 Log:0/13
 > 1 receiving messages

testdata/replicate_pause.txt

@@ -129,14 +129,14 @@ stabilize 1
   Ready MustSync=false:
   Messages:
   1->2 MsgHeartbeat Term:1 Log:0/0 Commit:17
-  1->3 MsgHeartbeat Term:1 Log:0/0 Commit:11
+  1->3 MsgHeartbeat Term:1 Log:0/0 Commit:17
 
 stabilize 2 3
 ----
 > 2 receiving messages
   1->2 MsgHeartbeat Term:1 Log:0/0 Commit:17
 > 3 receiving messages
-  1->3 MsgHeartbeat Term:1 Log:0/0 Commit:11
+  1->3 MsgHeartbeat Term:1 Log:0/0 Commit:17
 > 2 handling Ready
   Ready MustSync=false:
   Messages:

testdata/snapshot_succeed_via_app_resp.txt

@@ -64,14 +64,14 @@ process-ready 1
 Ready MustSync=false:
 Messages:
 1->2 MsgHeartbeat Term:1 Log:0/0 Commit:11
-1->3 MsgHeartbeat Term:1 Log:0/0
+1->3 MsgHeartbeat Term:1 Log:0/0 Commit:11
 
 # Iterate until no more work is done by the new peer. It receives the heartbeat
 # and responds.
 stabilize 3
 ----
 > 3 receiving messages
-  1->3 MsgHeartbeat Term:1 Log:0/0
+  1->3 MsgHeartbeat Term:1 Log:0/0 Commit:11
   INFO 3 [term: 0] received a MsgHeartbeat message with higher term from 1 [term: 1]
   INFO 3 became follower at term 1
 > 3 handling Ready