[
https://issues.apache.org/jira/browse/SOLR-18587?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18132263#comment-18132263
]
Nick Shanin edited comment on SOLR-18587 at 10/26/26, 8:33 AM:
---------------------------------------------------------------
I was able to build a patch for this, but I can't submit a PR.
The CoreAdmin fix is two small commits on the candidate, and they touch one
product line and one test line.
Product fix, f6209594a8b: in CoreContainer.unloadWithoutCoreOp, the unload
skips the "core failed to init" shortcut when the core is still loaded. A
failed reload can leave the previous core loaded. Before, unload took the
shortcut and never closed that core. The condition is now loadFailure != null
&& !solrCores.isLoaded(name), with a comment about the failed reload.
Test fix, 8b8c45ca0a7: testDeleteInstanceDirAfterCreateFailure now calls
useFactory(null) first. The test checks that the data directory is deleted on
unload. The default test directory factory, MockDirectoryFactory, never deletes
files, so the test needs a filesystem-backed factory.
The product fix is what makes the RC3 behaviour correct. The test fix only
makes the test able to see it. The matched-pair runs showed the test fails on
RC3 and passes with the product fix plus the test change.
AI-generated, human-reviewed.
was (Author: JIRAUSER314749):
I was able to build a patch, but I can't submit a PR.
The CoreAdmin fix is two small commits on the candidate, and they touch one
product line and one test line.
Product fix, f6209594a8b: in CoreContainer.unloadWithoutCoreOp, the unload
skips the "core failed to init" shortcut when the core is still loaded. A
failed reload can leave the previous core loaded. Before, unload took the
shortcut and never closed that core. The condition is now loadFailure != null
&& !solrCores.isLoaded(name), with a comment about the failed reload.
Test fix, 8b8c45ca0a7: testDeleteInstanceDirAfterCreateFailure now calls
useFactory(null) first. The test checks that the data directory is deleted on
unload. The default test directory factory, MockDirectoryFactory, never deletes
files, so the test needs a filesystem-backed factory.
The product fix is what makes the RC3 behaviour correct. The test fix only
makes the test able to see it. The matched-pair runs showed the test fails on
RC3 and passes with the product fix plus the test change.
AI-generated, human-reviewed.
> CoreAdminHandlerTest.testDeleteInstanceDirAfterCreateFailure fails repeatedly
> in the 10.1.0 RC3 smoke tests
> -----------------------------------------------------------------------------------------------------------
>
> Key: SOLR-18587
> URL: https://issues.apache.org/jira/browse/SOLR-18587
> Project: Solr
> Issue Type: Test
> Reporter: Nick Shanin
> Priority: Major
>
> h2. What was seen
> In five completed runs of the Solr 10.1.0 RC3 smoke tester (artifacts rev
> {{749712b71940c9f96d824c533f32f5c56dcd3924}}, JDK 21.0.12.1, aarch64), the
> build failed every time, and
> {{CoreAdminHandlerTest.testDeleteInstanceDirAfterCreateFailure}} failed in
> all five runs, across different random seeds.
> h2. Caveat
> All five runs were on one small ARM host (4 CPUs, 7 GB), so a host effect is
> not yet separated from a code or test cause. The failure repeats across seeds
> on that host; whether it reproduces elsewhere is part of what this ticket
> tracks.
> h2. Related
> SOLR-9433 covers the SolrCore clean-up path this test exercises. Related RC3
> smoke failures: RecoveryAfterSoftCommitTest is SOLR-18530 (fix in PR #5098)
> and TestCoordinatorRole.testNRTRestart is SOLR-16630 (fix in PR #5106);
> neither fix is in RC3.
> 🤖 _AI text above_ 🤖 _(reviewed by Nick Shanin)_
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
