etcdserver: Guarantee order of requested progress notifications

Progress notifications requested using ProgressRequest were sent
directly using the ctrlStream, which means that they could race
against watch responses in the watchStream.

This would especially happen when the stream was not synced - e.g. if
you requested a progress notification on a freshly created unsynced
watcher, the notification would typically arrive indicating a revision
for which not all watch responses had been sent.

This changes the behaviour so that v3rpc always goes through the watch
stream, using a new RequestProgressAll function that closely matches
the behaviour of the v3rpc code - i.e.

1. Generate a message with WatchId -1, indicating the revision for
   *all* watchers in the stream

2. Guarantee that a response is (eventually) sent

The latter might require us to defer the response until all watchers
are synced, which is likely as it should be. Note that we do *not*
guarantee that the number of progress notifications matches the number
of requests, only that eventually at least one gets sent.

Signed-off-by: Benjamin Wang <wachao@vmware.com>

server/etcdserver/api/v3rpc/watch.go

@@ -144,6 +144,10 @@ type serverWatchStream struct {
 	// records fragmented watch IDs
 	fragment map[mvcc.WatchID]bool
 
+	// indicates whether we have an outstanding global progress
+	// notification to send
+	deferredProgress bool
+
 	// closec indicates the stream is closed.
 	closec chan struct{}
 
@@ -173,6 +177,8 @@ func (ws *watchServer) Watch(stream pb.Watch_WatchServer) (err error) {
 		prevKV:   make(map[mvcc.WatchID]bool),
 		fragment: make(map[mvcc.WatchID]bool),
 
+		deferredProgress: false,
+
 		closec: make(chan struct{}),
 	}
 
@@ -359,10 +365,16 @@ func (sws *serverWatchStream) recvLoop() error {
 			}
 		case *pb.WatchRequest_ProgressRequest:
 			if uv.ProgressRequest != nil {
-				sws.ctrlStream <- &pb.WatchResponse{
-					Header:  sws.newResponseHeader(sws.watchStream.Rev()),
-					WatchId: clientv3.InvalidWatchID, // response is not associated with any WatchId and will be broadcast to all watch channels
+				sws.mu.Lock()
+				// Ignore if deferred progress notification is already in progress
+				if !sws.deferredProgress {
+					// Request progress for all watchers,
+					// force generation of a response
+					if !sws.watchStream.RequestProgressAll() {
+						sws.deferredProgress = true
+					}
 				}
+				sws.mu.Unlock()
 			}
 		default:
 			// we probably should not shutdown the entire stream when
@@ -430,11 +442,15 @@ func (sws *serverWatchStream) sendLoop() {
 				Canceled:        canceled,
 			}
 
-			if _, okID := ids[wresp.WatchID]; !okID {
-				// buffer if id not yet announced
-				wrs := append(pending[wresp.WatchID], wr)
-				pending[wresp.WatchID] = wrs
-				continue
+			// Progress notifications can have WatchID -1
+			// if they announce on behalf of multiple watchers
+			if wresp.WatchID != clientv3.InvalidWatchID {
+				if _, okID := ids[wresp.WatchID]; !okID {
+					// buffer if id not yet announced
+					wrs := append(pending[wresp.WatchID], wr)
+					pending[wresp.WatchID] = wrs
+					continue
+				}
 			}
 
 			mvcc.ReportEventReceived(len(evs))
@@ -465,6 +481,11 @@ func (sws *serverWatchStream) sendLoop() {
 				// elide next progress update if sent a key update
 				sws.progress[wresp.WatchID] = false
 			}
+			if sws.deferredProgress {
+				if sws.watchStream.RequestProgressAll() {
+					sws.deferredProgress = false
+				}
+			}
 			sws.mu.Unlock()
 
 		case c, ok := <-sws.ctrlStream: