[
https://issues.apache.org/jira/browse/FLINK-25872?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17521607#comment-17521607
]
Yanfei Lei commented on FLINK-25872:
------------------------------------
hi [~roman], I wrote a
document([https://docs.google.com/document/d/1KSFWc0gL7HkhC-JNrnsp06TLnsTmZOTHITQDcGMo0cI/edit?usp=sharing|https://docs.google.com/document/d/1KSFWc0gL7HkhC-JNrnsp06TLnsTmZOTHITQDcGMo0cI/edit?usp=sharing,])
about this
ticket[,|https://docs.google.com/document/d/1KSFWc0gL7HkhC-JNrnsp06TLnsTmZOTHITQDcGMo0cI/edit?usp=sharing,]
would you please take a review and give some advice?
{quote}I think registering *all* {{{}KeyedStateHandle{}}}s with the
{{SharedStateRegistry}} on recovery in {{CLAIM}} mode would also solve the
problem, wouldn't it?
The advantage is that JM wouldn't have to know anything about the changelog.
I think this is important and that's why I'd prefer such an approach.
{quote}
I think only `{{{}registering *all* {{{}KeyedStateHandle{}}}s with the
{{SharedStateRegistry}}{}}}` may not work as well, because the
{{discardState()}} of KeyedStateHandles are {*}not empty{*}, although all
{{KeyedStateHandles}} are registered to {{{}SharedStateRegistry{}}}, the state
would be discarded on checkpoint subsuming(maybe I overlooked something)?
> 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.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
