JAkutenshi commented on code in PR #4446:
URL: https://github.com/apache/ignite-3/pull/4446#discussion_r1871772817
##########
modules/replicator/src/main/java/org/apache/ignite/internal/replicator/ReplicaManager.java:
##########
@@ -1298,6 +1319,93 @@ private CompletableFuture<Boolean>
onPrimaryExpired(PrimaryReplicaEventParameter
return falseCompletedFuture();
}
+ // TODO: move to Replica
https://issues.apache.org/jira/browse/IGNITE-23750
+ private void registerFailoverCallback(TablePartitionId
replicationGroupId) {
+ CompletableFuture<Replica> replicaFuture =
replicaManager.replica(replicationGroupId);
+
+ // TODO: https://issues.apache.org/jira/browse/IGNITE-23753
+ // For now we assume that it's a possible situation to get null
future on node's stopp process, but it should be fixed with
+ // a correct and consistent replica's state.
+ if (replicaFuture == null) {
+ return;
+ }
+
+ replicaFuture.thenApply(electedPrimaryReplica -> {
+ onLeaderElectedFailoverCallback = (leaderNode, term) ->
changePeersAndLearnersAsyncIfPendingExists(
+ electedPrimaryReplica,
+ replicationGroupId,
+ term
+ );
+
+ return
electedPrimaryReplica.raftClient().subscribeLeader(onLeaderElectedFailoverCallback);
Review Comment:
I may be don't fully understend. In the code above
`changePeersAndLearnersAsyncIfPendingExists` *is* the callback. I did it as the
field because (yet) I need the lambda's object to unsubscribe.
Also this callback doesn't send the new request, just straightforward call
`changePeersAndLearners` inside.
But `TableManager` do it, and yes, in worse case `changePeersAndLearners`
will be called twice: from the new request sent from `TableManager` and the
call from callback on leader elected.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: notifications-unsubscr...@ignite.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
