lei-xia commented on a change in pull request #1066:
URL: https://github.com/apache/helix/pull/1066#discussion_r459669342
##########
File path:
helix-core/src/main/java/org/apache/helix/controller/stages/MessageDispatchStage.java
##########
@@ -78,7 +78,17 @@ protected void processEvent(ClusterEvent event,
MessageOutput messageOutput) thr
batchMessage(dataAccessor.keyBuilder(), messagesToSend, resourceMap,
liveInstanceMap,
manager.getProperties());
+ String expectedSession =
event.getAttribute(AttributeName.EVENT_SESSION.name());
+ // An early check for expected leader session. If the sessions don't
match, it means the
+ // controller lost leadership, then messages should not be sent and
pipeline should stop.
+ // This avoids potential double masters for a single partition.
+ if (expectedSession != null &&
!expectedSession.equals(manager.getSessionId())) {
+ throw new StageException(
+ "Controller: " + manager.getInstanceName() + " lost leadership!
Expected session: "
+ + expectedSession + ", actual: " + manager.getSessionId());
+ }
Review comment:
should we do this for all Stage? or just messageDispatchStage? If for
all stages, maybe we should put this check before where each stage was called.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
