[jira] Commented: (CASSANDRA-1289) GossipTimerTask stops running if an Exception occurs

["Jonathan Ellis (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CASSANDRA-1289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12889554#action_12889554 ] 

Jonathan Ellis commented on CASSANDRA-1289:
-------------------------------------------

committed w/ changes since it was simple:

uses .error instead of .warn

uses .error(message, exception) so the entire stack trace will be logged

> GossipTimerTask stops running if an Exception occurs
> ----------------------------------------------------
>
>                 Key: CASSANDRA-1289
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-1289
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>    Affects Versions: 0.6, 0.6.1, 0.6.2, 0.6.3
>            Reporter: Wade Simmons
>            Assignee: Brandon Williams
>             Fix For: 0.6.4
>
>         Attachments: 1289.txt
>
>
> The GossipTimerTask run() method has a try/catch around its body, but it re-throws all Exceptions as RuntimeExceptions. This causes the GossipTimerTask to no longer run (due to the way the underlying Java Timer implementation works), stopping the periodic gossip status checks.
> Combine this problem with a bug like CASSANDRA-757 (not yet fixed in 0.6.x) and you get into a state where the server keeps running, but gossip is no longer occurring, preventing node addition / removal from happening.
> I see two potential choices:
> 1) Log the error but don't re-throw it so that the GossipTimerTask will continue to run on its next interval.
> 2) Shutdown the server, since continuing to run without gossip subtly breaks other functionality / knowledge of other nodes.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.