Save config file and brocast the PONG when configEpoch changed (#1813)

This is somehow related with #974 and #1777. When the epoch changes,
we should save the configuration file and broadcast a PONG as much
as possible.

For example, if a primary down after bumping the epoch, its replicas
may initiate a failover, but the other primaries may refuse to vote
because the epoch of the replica has not been updated.

Or for example, for some reasons we bump the epoch, if the epoch
is not updated in time in the cluster, it may affect the judgment
of message staleness.

These broadcasts are expensive in large clusters, but none of these
seem high frequency so it should be fine.

---------

Signed-off-by: Binbin <[email protected]>

Author: enjoy-binbin

src/cluster_legacy.c

@@ -1938,7 +1938,7 @@ int clusterBumpConfigEpochWithoutConsensus(void) {
     if (myself->configEpoch == 0 || myself->configEpoch != maxEpoch) {
         server.cluster->currentEpoch++;
         myself->configEpoch = server.cluster->currentEpoch;
-        clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | CLUSTER_TODO_FSYNC_CONFIG);
+        clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | CLUSTER_TODO_FSYNC_CONFIG | CLUSTER_TODO_BROADCAST_ALL);
         serverLog(LL_NOTICE, "New configEpoch set to %llu", (unsigned long long)myself->configEpoch);
         return C_OK;
     } else {
@@ -2001,7 +2001,7 @@ void clusterHandleConfigEpochCollision(clusterNode *sender) {
     /* Get the next ID available at the best of this node knowledge. */
     server.cluster->currentEpoch++;
     myself->configEpoch = server.cluster->currentEpoch;
-    clusterSaveConfigOrDie(1);
+    clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | CLUSTER_TODO_FSYNC_CONFIG | CLUSTER_TODO_BROADCAST_ALL);
     serverLog(LL_NOTICE, "configEpoch collision with node %.40s (%s). configEpoch set to %llu", sender->name,
               sender->human_nodename, (unsigned long long)myself->configEpoch);
 }
@@ -4774,13 +4774,13 @@ void clusterFailoverReplaceYourPrimary(void) {
         }
     }
 
-    /* 3) Update state and save config. */
+    /* 3) Update state, note that we do not use CLUSTER_TODO_UPDATE_STATE since we want the node
+     * to update the cluster state ASAP after failover. */
     clusterUpdateState();
-    clusterSaveConfigOrDie(1);
 
-    /* 4) Pong all the other nodes so that they can update the state
+    /* 4) Save and fsync the config, and pong all the other nodes so that they can update the state
      *    accordingly and detect that we switched to primary role. */
-    clusterDoBeforeSleep(CLUSTER_TODO_BROADCAST_ALL);
+    clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | CLUSTER_TODO_FSYNC_CONFIG | CLUSTER_TODO_BROADCAST_ALL);
 
     /* 5) If there was a manual failover in progress, clear the state. */
     resetManualFailover();
@@ -4963,6 +4963,7 @@ void clusterHandleReplicaFailover(void) {
         /* Update my configEpoch to the epoch of the election. */
         if (myself->configEpoch < server.cluster->failover_auth_epoch) {
             myself->configEpoch = server.cluster->failover_auth_epoch;
+            clusterDoBeforeSleep(CLUSTER_TODO_SAVE_CONFIG | CLUSTER_TODO_FSYNC_CONFIG | CLUSTER_TODO_BROADCAST_ALL);
             serverLog(LL_NOTICE, "configEpoch set to %llu after successful failover",
                       (unsigned long long)myself->configEpoch);
         }