[
https://issues.apache.org/jira/browse/FLINK-20488?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Piotr Nowojski updated FLINK-20488:
-----------------------------------
Comment: was deleted
(was: Good idea with the 2. point. About the 1.
I don’t like the behaviour of changing UC to AC, as it would add complexity to
the system, without clear benefit to me. Or maybe there is some rare use case
which would benefit from such change, but all the other uses cases would suffer
(more complexity, more complicated code, more moving parts, more things to
document and understand). After all, if someone wants to limit the amount of
persisted data, he can just increase the timeout value.
Also if the motivation is to limit persisted data, a better and I think more
useful idea would be to add size limit for in-flight data before triggering
unaligned checkpoint. It would let users to control the amount of persisted
data much more strictly.)
> Supporting changes (UI, conf) for timing out aligned checkpoints
> ----------------------------------------------------------------
>
> Key: FLINK-20488
> URL: https://issues.apache.org/jira/browse/FLINK-20488
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing, Runtime / Configuration,
> Runtime / Web Frontend
> Affects Versions: 1.13.0
> Reporter: Roman Khachatryan
> Priority: Major
>
> A follow-up ticket after FLINK-19681 to address issues not directly related
> to checkpointing (see
> [discussion|https://github.com/apache/flink/pull/13827#discussion_r527794600]).
> In the UI, show checkpoint type for each subtask; on a checkpoint level
> display unaligned if at least one subtask did UC.
> That should ease debugging of the checkpointing issues.
>
> (also consider renaming alignment timeout option to subtask alignment timeout)
>
> Disabling propagation moved to FLINK-20548.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
