[
https://issues.apache.org/jira/browse/FLINK-25322?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17750150#comment-17750150
]
Hangxiang Yu commented on FLINK-25322:
--------------------------------------
{quote}But I don't think this adds complexity to the user, because before that
the user also has to check the Flink UI or REST UI to determine whether the new
job has completed at least one checkpoint.
{quote}
Just my intuitive, I think checking the status of first checkpoint is simpler
than checking a new added status for users.
BTW, It's also a bit strange if users set more than one retained checkpoints,
some checkpoints could be used, others could not because they may not be
NO_CLAIM which still have dependency with previous one.
{quote}Another option is to record in the checkpoint metadata which state
artifacts are borrowed from the non-claimed checkpoint, and when the new
checkpoint is used for claim mode recovery, those state artifacts borrowed from
the non-claimed checkpoint will not be deleted.
{quote}
I agree with [~pnowojski]. It behaves not like NO_CLAIM mode.
I also think it may be acceptable if just first checkpoint is a bit long for
changelog state backend in NO_CLAIM mode.
{quote}On this basis, what if we do not execute notifyCheckpointCompleted for
these not state self-sustained checkpoints ?
{quote}
IIUC, You mean that triggering the checkpoint commit but not making the
checkpoint complete ?
If true, I don't think it's a good idea. It also breaks previous checkpoint
behavior.
> 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)
