[
https://issues.apache.org/jira/browse/IGNITE-8422?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16467302#comment-16467302
]
Ivan Rakov commented on IGNITE-8422:
------------------------------------
[~Jokser], I've reviewed your changes. In general all looks good, but I have a
few comments regarding yellow code and code style:
1. BitSetIterator#next doesn't throw NoSuchElementException
2. ClusterGraph - unused import, abbreviations are not used
3. DefaultCommunicationFailureResolver#clusterNodeIds - block around one-liner
4. FullyConnectedComponentSearcher - abbreviations
5. FullyConnectedComponentSearcherTest - abbreviations, missing comments
> Zookeeper discovery split brain detection shouldn't consider client nodes
> -------------------------------------------------------------------------
>
> Key: IGNITE-8422
> URL: https://issues.apache.org/jira/browse/IGNITE-8422
> Project: Ignite
> Issue Type: Bug
> Components: zookeeper
> Affects Versions: 2.5
> Reporter: Pavel Kovalenko
> Assignee: Pavel Kovalenko
> Priority: Major
> Fix For: 2.6
>
>
> Currently Zookeeper discovery checks each splitted cluster on full
> connectivity taking into account client nodes. This is not correct, because
> server and client nodes may use different networks to connect to each other.
> It means that there can be client which sees both parts of splitted cluster
> and breaks split brain recovery - full connected part of server nodes will be
> never find.
> We should exclude client nodes from split brain analysis and improve split
> brain tests to make them truly fair.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
