[
https://issues.apache.org/jira/browse/FLINK-41285?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18145107#comment-18145107
]
Chanhae Oh commented on FLINK-41285:
------------------------------------
Hi,[~chenzihao], [~panyuepeng] .
I have completed the verification.
Please find the attached PDF and the shell script used to run the verification.
1. **Build**: `./mvnw clean install -DskipTests -Dfast -Djdk17 -Pjava17-target
-T1C -pl flink-dist -am` (webui included)
2. `./verify.sh setup` — generate cluster config + FILE/ROCKSDB HistoryServer
configs
3. `./verify.sh start-cluster` — start JobManager + TaskManager
4. `./verify.sh run-jobs` — run 4 example jobs: WordCount (FINISHED), AsyncIO
(FAILED), TopSpeedWindowing (CANCELED), WordCount with bad input path (FAILED)
5. `./verify.sh start-hs` — start both HistoryServer instances;
`/jobs/overview` and all per-job endpoints
(detail/vertices/config/exceptions/accumulators/checkpoints/plan) matched
byte-for-byte; local file count FILE=110 vs ROCKSDB=8
6. `./verify.sh retention` — set `retained-jobs/applications: 2`, restart; both
modes evicted the same 2 oldest entries identically
7. `./verify.sh restart-recovery` — restart HistoryServer; both modes recovered
the remaining jobs identically from the archive directory
Result: FILE and ROCKSDB modes are equivalent — identical REST responses, far
fewer local files, identical retention/cleanup behavior, identical recovery
after restart.
Thank you.
> Release Testing: Verify FLIP-584: Support Pluggable Storage Backend for
> HistoryServer
> -------------------------------------------------------------------------------------
>
> Key: FLINK-41285
> URL: https://issues.apache.org/jira/browse/FLINK-41285
> Project: Flink
> Issue Type: Sub-task
> Reporter: chenzihao
> Assignee: Chanhae Oh
> Priority: Blocker
> Labels: release-testing
> Fix For: 2.4.0
>
> Attachments: FLIP-584-verification.pdf, verify.sh
>
>
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)