hadoop-common-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Eli Collins (Commented) (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HADOOP-8236) haadmin should have configurable timeouts for failover commands
Date Sat, 31 Mar 2012 16:38:25 GMT

    [ https://issues.apache.org/jira/browse/HADOOP-8236?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13243220#comment-13243220
] 

Eli Collins commented on HADOOP-8236:
-------------------------------------

Todd,
These timeouts look reasonable to me. Worth noting that new-active is also the timeout for
the active pre-check, ie the check that the new active is alive and well before we ask the
current active to go standby. This is important because we don't want to impatiently wait
5s before fencing then wait a minute to make the new active active. In practice since we already
contacted the new active we probably won't have to wait 60s to transition it to active unless
something happened in between the pre-check and the transition to active, which is why 60s
timeout here is reasonable.

Nit: can remove the "TODO" before transitionToActive since this is now configurable.
Otherwise patch looks great.

                
> haadmin should have configurable timeouts for failover commands
> ---------------------------------------------------------------
>
>                 Key: HADOOP-8236
>                 URL: https://issues.apache.org/jira/browse/HADOOP-8236
>             Project: Hadoop Common
>          Issue Type: Improvement
>          Components: ha
>    Affects Versions: 0.23.3
>            Reporter: Philip Zeyliger
>            Assignee: Todd Lipcon
>         Attachments: hadoop-8236.txt
>
>
> The HAAdmin failover could should time out reasonably aggressively and go onto the fencing
strategies if it's dealing with a mostly dead active namenode.  Currently it uses what's probably
the default, which is to say no timeout whatsoever.
> {code}
>   /**
>    * Return a proxy to the specified target service.
>    */
>   protected HAServiceProtocol getProtocol(String serviceId)
>       throws IOException {
>     String serviceAddr = getServiceAddr(serviceId);
>     InetSocketAddress addr = NetUtils.createSocketAddr(serviceAddr);
>     return (HAServiceProtocol)RPC.getProxy(
>           HAServiceProtocol.class, HAServiceProtocol.versionID,
>           addr, getConf());
>   }
> {code}

--
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

        

Mime
View raw message