[
https://issues.apache.org/jira/browse/SOLR-12313?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16682055#comment-16682055
]
Hoss Man commented on SOLR-12313:
---------------------------------
bq. RecoveryAfterSoftCommitTest has been failing roughly 50% of the time the
past few days - but only on master, ...
actually, it just occurred to me i was thinking about it backwards: it's
failing almost 50% of the time it's run by jenkins, but ~ half of those are on
7x where this change wasn't made. Which means on master jenkins builds, it's
failing almost 100% of the time.
...
It looks like {{tests.asserts}} is the key variable -- any seed i try with
{{-Dtests.asserts=true}} seems to fail, but {{-Dtests.asserts=false}} passes.
> TestInjection#waitForInSyncWithLeader needs improvement.
> --------------------------------------------------------
>
> Key: SOLR-12313
> URL: https://issues.apache.org/jira/browse/SOLR-12313
> Project: Solr
> Issue Type: Test
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Mark Miller
> Priority: Major
>
> This really should have some doc for why it would be used.
> I also think it causes BasicDistributedZkTest to take forever for sometimes
> and perhaps other tests?
> I think checking for uncommitted data is probably a race condition and should
> be removed.
> Checking index versions should follow the rules that replication does - if
> the slave is higher than the leader, it's in sync, being equal is not
> required. If it's expected for a test it should be a specific test that
> fails. This just introduces massive delays.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
