[
https://issues.apache.org/jira/browse/CASSANDRA-3654?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13173588#comment-13173588
]
Brandon Williams commented on CASSANDRA-3654:
---------------------------------------------
Can we mention this ticket in the patch so if people do encounter this they
have an easier avenue to pursue?
I'll go ahead and add a little note here quoting your blog post:
bq. I think in theory there is a danger that if we removed node #23 from the
ring and added another node with the same IP in less than 4 days it will have
problems with it’s Generation not been seen as new. But thats a pretty small
danger.
Yes, that would definitely be a problem, because removed states are held until
Gossiper.aVeryLongTime (after CASSANDRA-2496) which is 3 days (I quoted you 4
just to be safe.) If the removed generation is in the future and thus higher
than the joining node, the new node won't be able to communicate any state for
either that IP *or* that token. The easiest thing to do in this situation is
bootstrap the new node on a different IP at token-1, removetoken the existing
token, wait 3 days, and if you really care about having the old IP or token
back move it after that. You can also use this as an alternate solution to the
problem, though you _must_ wipe at least the system keyspace on the node before
bootstrapping it back in.
> Warn when the stored Gossip Generation is from the future
> ---------------------------------------------------------
>
> Key: CASSANDRA-3654
> URL: https://issues.apache.org/jira/browse/CASSANDRA-3654
> Project: Cassandra
> Issue Type: Improvement
> Components: Core
> Affects Versions: 1.0.6
> Reporter: Aaron Morton
> Assignee: Aaron Morton
> Priority: Trivial
> Attachments: 0001-3654.patch
>
>
> I had a case where the server was first started with the current time set way
> in the future. So the gossip generation was initialized with a very high
> value (background
> http://thelastpickle.com/2011/12/15/Anatomy-of-a-Cassandra-Partition/)
> There were some other issues at play, but a log message warning of the high
> generation would have helped.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
