rkhachatryan commented on a change in pull request #15728:
URL: https://github.com/apache/flink/pull/15728#discussion_r621272696
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/DefaultCheckpointPlanCalculator.java
##########
@@ -111,7 +112,10 @@ public void setAllowCheckpointsAfterTasksFinished(boolean
allowCheckpointsAfterT
? calculateAfterTasksFinished()
: calculateWithAllTasksRunning();
- checkTasksStarted(result.getTasksToTrigger());
+ checkTasksStarted(
+ isUnalignedCheckpoint
+ ? result.getTasksToWaitFor()
+ : result.getTasksToTrigger());
Review comment:
@akalash
> Do you mean 'checkpoint timeout must be greater than recovery time'? If I
understand checkpoint timeout correctly, then the checkpoint will be discarded
and it will start from beginning if checkpoint timeout is less than recovery
time.
Yes (it's a matter of wording).
Timeouts are counted as checkpoint failures, and if this count reaches a
threshold then the job will fail.
@pnowojski
> Currently first checkpoint would complete after 1h 10 minutes. If we wait
for recovery to complete before triggering first checkpoint, first checkpoint
would complete after 2h.
Given that the checkpoint timeout is big enough (default is 10m, for larger
setups I saw 30m).
@AHeise
> That is, I'm assuming both ways are fine with their disadvantages and we
should just pick one. In doubt, pick the one that preserves the old behavior.
In this case old behavior (for AC) is combined with the new behaviour (for
UC) and this seems misleading from both users and developers point of view IMO.
--
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]
