[
https://issues.apache.org/jira/browse/SOLR-8837?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15192317#comment-15192317
]
Alan Woodward commented on SOLR-8837:
-------------------------------------
We can fix this by fixing the duplicate-detection logic, but I'm not sure we
shouldn't just remove this entirely. I guess it could prevent an election
delay in the case that a session gets interrupted and the ephemeral nodes from
the previous session hang around up to the session timeout, but it seems just
as likely to cause problems to me.
> Duplicate leader elector node detection is broken
> -------------------------------------------------
>
> Key: SOLR-8837
> URL: https://issues.apache.org/jira/browse/SOLR-8837
> Project: Solr
> Issue Type: Bug
> Affects Versions: master, 6.0
> Reporter: Alan Woodward
>
> LeaderElector.checkIfIAmLeader checks to see if it has duplicate
> registrations in under its election node, but it does this by prefix
> checking, which means that if core_node1 registers itself after core_node11,
> it will think it's a duplicate, and delete the core_node11 node.
> This is causing regular failures in UnloadDistributedZkTest.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
