[jira] [Comment Edited] (FLINK-25872) Restoring from non-changelog checkpoint with changelog state-backend enabled in CLAIM mode discards state in use

["Piotr Nowojski (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/FLINK-25872?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17530927#comment-17530927 ] 

Piotr Nowojski edited comment on FLINK-25872 at 5/2/22 7:47 PM:
----------------------------------------------------------------

[~dwysakowicz] raised a good point. Do we really need a support for the switch in this direction: changelog off -> on? This seems like a planned use case, so it should be fine if savepoint takes more time. The other direction might be in order to avoid some infinite bug loops, so on -> off support from checkpoint makes more sense to me.  


was (Author: pnowojski):
I would second [~dwysakowicz] concerns if we really need a support for the switch in this direction: changelog off -> on. This seems like a planned use case, so it should be fine if savepoint takes more time. The other direction might be in order to avoid some infinite bug loops, so on -> off support from checkpoint makes more sense to me.  

> Restoring from non-changelog checkpoint with changelog state-backend enabled in CLAIM mode discards state in use
> ----------------------------------------------------------------------------------------------------------------
>
>                 Key: FLINK-25872
>                 URL: https://issues.apache.org/jira/browse/FLINK-25872
>             Project: Flink
>          Issue Type: Bug
>          Components: Runtime / Checkpointing, Runtime / State Backends
>            Reporter: Yun Tang
>            Assignee: Yanfei Lei
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 1.16.0
>
>
> If we restore from checkpoint with changelog state-backend enabled in snapshot CLAIM mode, the restored checkpoint would be discarded on subsume. This invalidates newer/active checkpoints because their materialized part is discarded (for incremental wrapped checkpoints, their private state is discarded). This bug is like FLINK-25478.
>  
> Design doc: [https://docs.google.com/document/d/1KSFWc0gL7HkhC-JNrnsp06TLnsTmZOTHITQDcGMo0cI/edit?usp=sharing,]



--
This message was sent by Atlassian Jira
(v8.20.7#820007)