Re: Slow flink checkpoint
Posted by
Fabian Hueske-2 on
URL: http://deprecated-apache-flink-user-mailing-list-archive.369.s1.nabble.com/Slow-flink-checkpoint-tp18946p19536.html
Hi everybody,
Thanks so much for looking into this issue and posting the detailed description of your approach.
As said before, improving the checkpointing performance for timers is a very important improvement for Flink.
I'm not familiar with the internals of the timer service checkpointing, but adding an add and delete version field and perform async checkpoints based on these fields seems like a good approach to me.
IIRC, Aljoscha (in CC) implemented the timer service and its checkpointing. He might have more comments.
I'd suggest to create a JIRA (everybody can do that) and repost the description of your approach there.
If you have the code ready, you can also open a PR and reference the JIRA.
Best, Fabian