1996fanrui commented on code in PR #21131:
URL: https://github.com/apache/flink/pull/21131#discussion_r1009238931
##########
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/channel/ChannelStateWriteRequestDispatcherImpl.java:
##########
@@ -98,18 +120,26 @@ private ChannelStateCheckpointWriter
buildWriter(CheckpointStartRequest request)
streamFactoryResolver.resolveCheckpointStorageLocation(
request.getCheckpointId(),
request.getLocationReference()),
serializer,
- () -> writers.remove(request.getCheckpointId()));
+ () -> {
+ checkState(
+ request.getCheckpointId() == ongoingCheckpointId,
+ "The ongoingCheckpointId[%s] was changed when
clear writer of checkpoint[%s], it might be a bug.",
+ ongoingCheckpointId,
+ request.getCheckpointId());
+ this.writer = null;
+ });
Review Comment:
Hi @pnowojski
As I mentioned in the last comment, `writer` cannot be set to `null` during
abort checkpoint.
> If the 2. happens after 1., this checkState would fail, right?
I don't think so. Actually, all `abort`, `start`, `writeInput`,
`finishInput`, `writeOutput` or `finishOutput` are executed in the `Channel
state writer Executor` Thread. It is single thread, so the race condition isn't
exist.
Please correct me if anything is wrong.
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
