[
https://issues.apache.org/jira/browse/SOLR-8249?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14994425#comment-14994425
]
Ishan Chattopadhyaya commented on SOLR-8249:
--------------------------------------------
The OverseerTest's testOverseerStatsReset() method sets up Zk, overseer, lets
an election happen and at the end of it counts the number of STATE messages
sent.
In SOLR-7989, one extra STATE message is passed in if the replica chosen as the
leader was not active. This breaks the test at Jenkins. However, I was not able
to reproduce it when I was testing before SOLR-7989 was committed; the reason
might be either (a) the replica was active, and hence no extra STATE message
was sent, or (b) the assert for STATE messages is being done before the leader
election is happening (timing issue).
I propose that we change the test in the following two ways: (a) before the
asserts, wait a while for leader election to have happened. (b) instead of
checking for number of STATE messages, we can check for number of LEADER
messages.
> OverseerTest failures
> ---------------------
>
> Key: SOLR-8249
> URL: https://issues.apache.org/jira/browse/SOLR-8249
> Project: Solr
> Issue Type: Bug
> Reporter: Noble Paul
>
> http://jenkins.thetaphi.de/job/Lucene-Solr-trunk-Windows/5377/
> This is related to SOLR-7989 . [~ichattopadhyaya] we need to fix the testcase
--
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
