[jira] Issue Comment Edited: (CASSANDRA-1894) Consistency manager is unbounded

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

    [ https://issues.apache.org/jira/browse/CASSANDRA-1894?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12974419#action_12974419 ] 

Jonathan Ellis edited comment on CASSANDRA-1894 at 12/22/10 5:35 PM:
---------------------------------------------------------------------

Requests don't back up in ConsistencyChecker the way they used to in mutation or read stages: it's already bounded to exactly RPC_TIMEOUT's worth of activity.

      was (Author: jbellis):
    Requests don't back up in ConsistencyChecker the way they do in mutation or read stages: it's already bounded to exactly RPC_TIMEOUT's worth of activity.
  
> Consistency manager is unbounded
> --------------------------------
>
>                 Key: CASSANDRA-1894
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-1894
>             Project: Cassandra
>          Issue Type: Bug
>    Affects Versions: 0.6.8
>            Reporter: Brandon Williams
>            Priority: Minor
>             Fix For: 0.6.9
>
>
> If a lot of reads are being performed at CL.ONE, and a replica node is 'sick' (that is, dying or going down, but the FD has not marked it down yet) then an unbounded amount of RR requests can back up in the consistency manager.  It is probably better to cap this queue, and drop further RR requests, rather than allow it to exert GC pressure.

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