[
https://issues.apache.org/jira/browse/SOLR-11426?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16198871#comment-16198871
]
Erick Erickson commented on SOLR-11426:
---------------------------------------
OK, no failures on master since this latest info-gathering code was committed
to master, which is kind of scary as this isn't a proper fix at all.
All the latest test does is fire a commit whenever a transient core is aged
out. I expected that just the normal core closing would issue a commit, and it
seems to, just not all the time apparently Which means there's some
circumstance where a core could lose documents that haven't been committed when
it is closed.
So the underlying issue is (apparently) there is some condition where
uncommitted changes are not committed on a core close. I'll be looking some
more, but would certainly welcome any ideas/help.
Meanwhile I'll leave the code in for a bit just to see if I'm getting (un)lucky
with this test and happening to succeed for a few days. If it continues to
succeed through tonight I'll revert tomorrow and insure it comes back.
> TestLazyCores fails too often
> -----------------------------
>
> Key: SOLR-11426
> URL: https://issues.apache.org/jira/browse/SOLR-11426
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Erick Erickson
> Assignee: Erick Erickson
>
> Rather then re-opening SOLR-10101 I thought I'd start a new issue. I may have
> to put some code up on Jenkins to test, last time I tried to get this to fail
> locally I couldn't
--
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
