[
https://issues.apache.org/jira/browse/SOLR-11168?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
David Smiley updated SOLR-11168:
--------------------------------
Attachment: output.txt
(A) I noticed the test contains:
{code:java}
} catch (Throwable t) {
t.printStackTrace();
}
{code}
which explains why the test threads that add documents report this
java.util.concurrent.RejectedExecutionException a bunch of times instead of say
once/thread. I think we could simply wrap non-runtime exceptions to throw via
Guava Throwables. _Nonetheless, this certainly isn't the bug._
(B) The test calls threadPool.awaitTermination with 5 seconds but there is no
code before that will actually terminate this threadpool, and so practically
speaking this will always take 5 seconds. Their is a shutdown() call after but
it should be before. _Nonetheless, this is probably not the bug._
(C) Maybe 5 seconds isn't enough time? We could add more time. Whatever time
we choose, if we detect that the threads are still busy, we might as well fail
right then and there with a helpful error message, rather then continue on and
assert things that will probably fail due to the CUSC being shut down
(RejectedExecution).
(D) CUSC.shutdownNow is being called... FWIW I think the code inside the
shutdownExecutor condition is probably not ideal. Seeing shutdownNow get called
multiple times is suspicious. And the shutdown then nearly immediate
shutdownNow following it is suspicious.
CC [~markrmil...@gmail.com]
> ConcurrentUpdateSolrClientTest failure, RejectedExecutionException
> ------------------------------------------------------------------
>
> Key: SOLR-11168
> URL: https://issues.apache.org/jira/browse/SOLR-11168
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Components: clients - java
> Affects Versions: 7.0
> Reporter: David Smiley
> Attachments: output.txt
>
>
> Jenkins failed https://builds.apache.org/job/Lucene-Solr-SmokeRelease-7.0/17/
> from test failure of ConcurrentUpdateSolrClientTest.
> I'm attaching the output for this test failure. It does not reproduce for me
> but that's not surprising as it seems this is a timing/concurrency issue.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
