[
https://issues.apache.org/jira/browse/CASSANDRA-1895?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12975881#action_12975881
]
Nick Bailey commented on CASSANDRA-1895:
----------------------------------------
I just noticed that the logs on one of the machines contains a "Removing token"
line followed by a new node line 3 times. Since Stu only called loadbalance
once, I believe this is likely due to gossip issues, where a node completes the
decom then gets gossip about the same node leaving and readds it.
I've also talked to one other person in irc who had a similar problem. Every
time a decommission happened, the node would get removed but added 30 seconds
later (gossip timeout after removal) due to some other node in the cluster.
> Loadbalance during gossip issues leaves cluster in bad state
> ------------------------------------------------------------
>
> Key: CASSANDRA-1895
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1895
> Project: Cassandra
> Issue Type: Bug
> Affects Versions: 0.8
> Reporter: Stu Hood
> Assignee: Stu Hood
> Fix For: 0.8
>
> Attachments: logs.tgz, ring-views.txt
>
>
> Running loadbalance against a node in a 4 node cluster leaves gossip in a
> wonky state.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
