pnowojski commented on a change in pull request #12186:
URL: https://github.com/apache/flink/pull/12186#discussion_r426117545
##########
File path:
flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/tasks/SubtaskCheckpointCoordinatorImpl.java
##########
@@ -174,7 +174,9 @@ public void notifyCheckpointComplete(long checkpointId,
OperatorChain<?, ?> oper
LOG.debug("Notification of complete checkpoint for task
{}", taskName);
for (StreamOperatorWrapper<?, ?> operatorWrapper :
operatorChain.getAllOperators(true)) {
-
operatorWrapper.getStreamOperator().notifyCheckpointComplete(checkpointId);
+ if (!operatorWrapper.isClosed()) {
Review comment:
I wonder if we shouldn't throw some exception?
##########
File path:
flink-streaming-java/src/main/java/org/apache/flink/streaming/runtime/tasks/SubtaskCheckpointCoordinatorImpl.java
##########
@@ -279,16 +281,18 @@ private void takeSnapshotSync(
try {
for (StreamOperatorWrapper<?, ?> operatorWrapper :
operatorChain.getAllOperators(true)) {
- operatorSnapshotsInProgress.put(
-
operatorWrapper.getStreamOperator().getOperatorID(),
- buildOperatorSnapshotFutures(
- checkpointMetaData,
- checkpointOptions,
- operatorChain,
-
operatorWrapper.getStreamOperator(),
- isCanceled,
- channelStateWriteResult,
- storage));
+ if (!operatorWrapper.isClosed()) {
Review comment:
ditto about exception or letting checkpoint coordinator know, that this
checkpoint was declined. Otherwise, aren't we risking a situation where this
checkpoint will complete, despite this operator not participating in it?
Also it might be better to move this if check much higher, somewhere to the
top of `checkpointState` method?
----------------------------------------------------------------
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]
