[
https://issues.apache.org/jira/browse/SOLR-7956?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Scott Blum updated SOLR-7956:
-----------------------------
Attachment: SOLR-7956-commit-tracker.patch
Some of those might be okay, I don't know the code well enough. However, the
shutdownNow() call in CommitTracker is definitely fatal. I've attached a patch
to demonstrate with a test case.
FWIW, the pending.cancel(true), while dangerous, is far less likely to be hit
because the task nulls out the pending field as soon as it starts to run. But
the shutdownNow() will interrupt the thread even after the field is nulled out,
so it's a lot more likely to really hit.
> There are interrupts on shutdown in places that can cause
> ChannelAlreadyClosed exceptions which prevents proper closing of transaction
> logs.
> --------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: SOLR-7956
> URL: https://issues.apache.org/jira/browse/SOLR-7956
> Project: Solr
> Issue Type: Bug
> Reporter: Mark Miller
> Assignee: Mark Miller
> Fix For: Trunk, 5.4
>
> Attachments: SOLR-7956-commit-tracker.patch, SOLR-7956.patch,
> SOLR-7956.patch, SOLR-7956.patch
>
>
> Found this while beast testing HttpPartitionTest.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
