[
https://issues.apache.org/jira/browse/HUDI-7527?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ethan Guo updated HUDI-7527:
----------------------------
Description: When the write client sends the request to the timeline server
with the latest instant and timeline hash, the latest instant and timeline hash
are based on the filtered timeline that does not contain pending log compaction
or clustering (replacecommit). In such a case, consecutive async scheduling of
clustering, even with a newly instantiated write client, does not trigger FSV
refresh at the timeline server; but FSV is affected by pending clustering and
log compaction (e.g., fgIdToPendingLogCompaction, fgIdToReplaceInstants,
fgIdToPendingClustering).
> Include instants outside commits and compaction for generating the latest
> instant and timeline hash for timeline server requests
> --------------------------------------------------------------------------------------------------------------------------------
>
> Key: HUDI-7527
> URL: https://issues.apache.org/jira/browse/HUDI-7527
> Project: Apache Hudi
> Issue Type: Bug
> Reporter: Ethan Guo
> Assignee: Ethan Guo
> Priority: Blocker
> Fix For: 0.15.0, 1.0.0
>
>
> When the write client sends the request to the timeline server with the
> latest instant and timeline hash, the latest instant and timeline hash are
> based on the filtered timeline that does not contain pending log compaction
> or clustering (replacecommit). In such a case, consecutive async scheduling
> of clustering, even with a newly instantiated write client, does not trigger
> FSV refresh at the timeline server; but FSV is affected by pending clustering
> and log compaction (e.g., fgIdToPendingLogCompaction, fgIdToReplaceInstants,
> fgIdToPendingClustering).
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
