[
https://issues.apache.org/jira/browse/FLINK-22008?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17310578#comment-17310578
]
xiaogang zhou commented on FLINK-22008:
---------------------------------------
if a metadata is invalid, someone search the directory and recover from it
will fail. if we write to a temp file and commit it will make it more dependent
> writing metadata is not an atomic operation, we should add a commit logic
> -------------------------------------------------------------------------
>
> Key: FLINK-22008
> URL: https://issues.apache.org/jira/browse/FLINK-22008
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Checkpointing
> Affects Versions: 1.12.2
> Reporter: xiaogang zhou
> Priority: Major
> Labels: pull-request-available
>
> writing metadata is not an atomic operation, some logic can cause there is a
> metadata file in the checkpoint dir, but the data is corrupted if the
> jobmanager crash while writing the metadata.
>
> So we should consider to add commit operation in the checkpoint storage
> location
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
