Merge pull request #5907 from Vincent-lau/private/shul2/coro-wait

CA-396635: Wait for corosync to update its info

ocaml/xapi/xapi_clustering.ml

@@ -545,6 +545,13 @@ module Watcher = struct
      is an update. *)
   let cluster_change_interval = Mtime.Span.(5 * min)
 
+  (* CA-396635: Sometimes it takes the underlying cluster stack (corosync) some time
+     to return a consistent view of the quorum. For example, it may be that the membership
+     information correctly reflects the new members after a membership change, while the
+     quorum field is still out of date. Add a delay here to make sure that the information
+     from corosync represents a consistent snapshot of the current cluster state. *)
+  let stabilising_period = Mtime.Span.(5 * s)
+
   let cluster_stack_watcher : bool Atomic.t = Atomic.make false
 
   (* we handle unclean hosts join and leave in the watcher, i.e. hosts joining and leaving
@@ -558,18 +565,20 @@ module Watcher = struct
           "cluster change watcher call"
           (Clock.Timer.span_to_s cluster_change_interval)
       in
-      let find_cluster_and_update updates =
+      let find_cluster_and_update ?(wait = false) updates =
         match find_cluster_host ~__context ~host with
         | Some ch ->
             let cluster = Db.Cluster_host.get_cluster ~__context ~self:ch in
+            if wait then
+              Thread.delay (Clock.Timer.span_to_s stabilising_period) ;
             on_corosync_update ~__context ~cluster updates
         | None ->
             ()
       in
       match Idl.IdM.run @@ Cluster_client.IDL.T.get m with
       | Ok updates ->
           (* Received updates from corosync-notifyd *)
-          find_cluster_and_update updates
+          find_cluster_and_update ~wait:true updates
       | Error (InternalError "UPDATES.Timeout") ->
           (* UPDATES.get timed out, this is normal.  *)
           (* CA-395789: We send a query to xapi-clusterd to fetch the latest state