hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Todd Lipcon (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-2465) HMaster should not contact each RS on startup
Date Mon, 19 Apr 2010 05:50:50 GMT

    [ https://issues.apache.org/jira/browse/HBASE-2465?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12858386#action_12858386
] 

Todd Lipcon commented on HBASE-2465:
------------------------------------

Essentially I have an RPC deadlock. The regionservers are sitting in reportForDuty trying
to report to the master, whose RPC server isn't up. The master is sitting in verifyClusterState
trying to talk to the regionservers, whose RPC servers aren't up.

I think the fix is that one of the two sides needs to run its RPC server before trying to
contact the other.

> HMaster should not contact each RS on startup
> ---------------------------------------------
>
>                 Key: HBASE-2465
>                 URL: https://issues.apache.org/jira/browse/HBASE-2465
>             Project: Hadoop HBase
>          Issue Type: Improvement
>          Components: master
>            Reporter: Todd Lipcon
>
> On startup, in verifyClusterState, the master contacts each region server serially. If
a region server is down it will retry for several minutes (if the client retry setting is
high). During this period, the master cannot be shut down, and also isn't processing real
work.

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


Mime
View raw message