cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Sylvain Lebresne (JIRA)" <>
Subject [jira] [Commented] (CASSANDRA-5167) Node isn't removed from system.peers after 'nodetool removenode'
Date Fri, 18 Jan 2013 09:38:13 GMT


Sylvain Lebresne commented on CASSANDRA-5167:

bq. I'm not sure why we ever made a 'removeTokens' method instead of 'removeEndpoint'

That slightly confused me too in CASSANDRA-4351, though I went for the lazy option of maintaining
behavior blindly.

So definitely agreed on the idea. Two tiny remark on the patch though:
* in SS.handleStateRemoving, we call Gossiper.instance.removeEndpoint() but shouldn't we called
SystemTable.removeEndpoint() too? In fact, it seems to me we may want to group both in a small
utility method since at least when we call the Gossiper one, I think we always should call
SystemTable one.
* in SS.handleStateNormal, ST.removeEndpoint should probably only be call if {{!isClient}}
like in SS.excise().
> Node isn't removed from system.peers after 'nodetool removenode'
> ----------------------------------------------------------------
>                 Key: CASSANDRA-5167
>                 URL:
>             Project: Cassandra
>          Issue Type: Bug
>    Affects Versions: 1.2.0
>         Environment: Ubuntu 12.10, Java 1.7.0_09 (OpenJDK)
>            Reporter: Nicolai Gylling
>            Assignee: Brandon Williams
>            Priority: Minor
>             Fix For: 1.2.1
>         Attachments: 5167.txt
> In a 3 node live cluster - After a replacement of a dead node, the old node remains in
the system.peers table, even after running 'nodetool removenode <ID>'.

This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see:

View raw message