[
https://issues.apache.org/jira/browse/FLINK-21352?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17457864#comment-17457864
]
Yuan Mei commented on FLINK-21352:
----------------------------------
Hey, [~roman] let's finalize things in this umbrella ticket that complete the
MVP (and get into 1.15), and maybe close/move out tickets as follow-ups for the
1.15?
1. We can remove/close all the "state ownership" tickets, and I think the
design should be related to FLIP 193 snapshot ownership as well. So we can
recreate/reopen them when needed.
2. We can keep all the test tickets.
3. Benchmarks can be kept but as a best effort.
Codewise needs to get in
https://issues.apache.org/jira/browse/FLINK-24611
Prevent JM discarding states when checkpoint abort is the only PR
Things are needed:
1. Support recovery (from checkpoints) after disabling the changelog backend
(if we do not have this, this feature is not safe to use in production).
Things I am not sure to include or not:
1. State migration
If we can make sure Savepoint does not use changelog, I think we should be fine
to leave it as it is. (I remember we have a randomization test that covered
this right?)
2. Consider disallowing in-memory state handles for materialized state
3. Support more than one retained checkpoints
> FLIP-158: Generalized incremental checkpoints
> ---------------------------------------------
>
> Key: FLINK-21352
> URL: https://issues.apache.org/jira/browse/FLINK-21352
> Project: Flink
> Issue Type: New Feature
> Components: Runtime / Checkpointing, Runtime / State Backends
> Reporter: Roman Khachatryan
> Assignee: Roman Khachatryan
> Priority: Major
> Fix For: 1.15.0
>
>
> Umbrella ticket for [FLIP-158: Generalized incremental
> checkpoints|https://cwiki.apache.org/confluence/display/FLINK/FLIP-158%3A+Generalized+incremental+checkpoints]
>
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
