[
https://issues.apache.org/jira/browse/HUDI-3406?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17501968#comment-17501968
]
Raymond Xu commented on HUDI-3406:
----------------------------------
[~shivnarayan] 2nd thought on the logic: is it actually better to always use fs
listing for rollback operation? 1st using commit metadata saves fs load but i
don't think perf is a big concern given rollback is not a frequent event. 2nd
fs listing serves a better source of truth for files to be cleaned up. can we
make a call here? cc [~x1q1j1] [~alexey.kudinkin]
> Rollback incorrectly relying on FS listing instead of Commit Metadata
> ---------------------------------------------------------------------
>
> Key: HUDI-3406
> URL: https://issues.apache.org/jira/browse/HUDI-3406
> Project: Apache Hudi
> Issue Type: Bug
> Reporter: Alexey Kudinkin
> Assignee: Forward Xu
> Priority: Blocker
> Labels: pull-request-available
> Fix For: 0.11.0
>
>
> [ListingBasedRollbackHelper|https://github.com/apache/hudi/blob/973087f385f0e42bd42e2d57008a756e182041ff/hudi-client/hudi-client-common/src/main/java/org/apache/hudi/table/action/rollback/ListingBasedRollbackHelper.java#L135]
> is relying on FS listing (`getBaseAndLogFilesToBeDeleted`) to determine
> which files should be deleted instead of relying on Commit Metadata for
> successfully completed commits.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
