[
https://issues.apache.org/jira/browse/HUDI-5044?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ethan Guo updated HUDI-5044:
----------------------------
Description: When clustering is disabled after being enabled, the inflight
clustering replacecommit scheduled and executed before will stay in the active
timeline indefinitely, blocking the metadata table compaction and data table
archival. We should add a logic to roll back inflight clustering replacecommit
when clustering is disabled. This should be guarded by a feature flag so if
clustering has never been enabled, there is no need to exercise the logic which
can introduce overhead of checking clustering intants. (was: When clustering
is disabled after being enabled, the inflight clustering replacecommit
scheduled and executed before will stay in the active timeline indefinitely,
blocking the metadata table compaction and data table archival. We should add
a logic to )
> Roll back inflight clustering replacecommit when clustering is disabled
> -----------------------------------------------------------------------
>
> Key: HUDI-5044
> URL: https://issues.apache.org/jira/browse/HUDI-5044
> Project: Apache Hudi
> Issue Type: Improvement
> Reporter: Ethan Guo
> Priority: Major
> Fix For: 0.13.0
>
>
> When clustering is disabled after being enabled, the inflight clustering
> replacecommit scheduled and executed before will stay in the active timeline
> indefinitely, blocking the metadata table compaction and data table archival.
> We should add a logic to roll back inflight clustering replacecommit when
> clustering is disabled. This should be guarded by a feature flag so if
> clustering has never been enabled, there is no need to exercise the logic
> which can introduce overhead of checking clustering intants.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
