Github user bjlovegithub commented on the issue:
https://github.com/apache/flink/pull/2629
Maybe this is a solution. We will not think about changing the type of
`checkpointLock`, which is an `Object` ans quite efficient, and not change the
order of `broadcastBarriers` and `operator.snapshotState()`. By placing
**pause** `EmitterThread` codes in `StreamTask.performCheckpoint()`, like this:
` private boolean performCheckpoint(CheckpointMetaData
checkpointMetaData) throws Exception {
LOG.debug("Starting checkpoint {} on task {}",
checkpointMetaData.getCheckpointId(), getName());
synchronized (lock) {
if (isRunning) {
// stop working threads first.
for (StreamOperator operator :
operatorChain.getAllOperators()) {
if (operator instanceof
AsyncWaitOperator) {
operator.pauseEmitterThread();
}
}
// broadcast barriers after snapshot operators'
states.
operatorChain.broadcastCheckpointBarrier(
checkpointMetaData.getCheckpointId(), checkpointMetaData.getTimestamp()
);
checkpointState(checkpointMetaData);
return true;
} else {
return false;
}
}
}`
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
