[
https://issues.apache.org/jira/browse/FLINK-3558?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Stephan Ewen closed FLINK-3558.
-------------------------------
Resolution: Invalid
This issue is not valid any more. Checkpoints are no longer triggered fast on
startup, to compensate for "not yet running tasks".
> Many declined checkpoints on startup
> ------------------------------------
>
> Key: FLINK-3558
> URL: https://issues.apache.org/jira/browse/FLINK-3558
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
> Reporter: Ufuk Celebi
> Priority: Major
>
> During testing, I've noticed that the first successful checkpoint had a very
> high ID (> 2000). Looking at the logs, I've noticed that the initial
> checkpoints were all declined.
> After receiving a decline message, a new checkpoint is triggered immediately,
> which can lead to a large number declined checkpoints on startup:
> https://gist.github.com/anonymous/0ef0923b06be96bb5eaa
> We can investigate whether it is a better strategy to just wait for the next
> periodic checkpoint trigger.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
