[
https://issues.apache.org/jira/browse/FLINK-25322?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17738073#comment-17738073
]
Piotr Nowojski commented on FLINK-25322:
----------------------------------------
Hi [~Feifan Wang] thanks for picking up this topic.
Another idea might be to just use the native savepoint code path for this?
Unless it's also not supported, but then we could fix both missing features.
Regarding the performance, I wouldn't worry about it too much. The exact same
issue has RocksDB. The solution is to:
* use claim mode for recovery/restarts
* use/implement fast duplicating FS
Use cases that are not solved by either of the above, I think users can live
with them at the moment. We could improve this more, but I doubt it's worth the
effort right now. Note that so far nobody has complained about no-claim mode
not working for the changelog statebackend so far, so we shouldn't pre-maturely
optimise this. I would go with full eager/sync materialization.
{quote}
I think that in the no-claim mode, it is not necessary to enter the
state-sustained state at the first checkpoint, but to enter the state-sustained
state as soon as possible.
{quote}
That would require a bigger discussion I think. So I would prefer to not do
that unless we really have to.
> Support no-claim mode in changelog state backend
> ------------------------------------------------
>
> Key: FLINK-25322
> URL: https://issues.apache.org/jira/browse/FLINK-25322
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Checkpointing, Runtime / State Backends
> Reporter: Dawid Wysakowicz
> Assignee: Feifan Wang
> Priority: Major
> Fix For: 1.18.0
>
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
