You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@hbase.apache.org by "Jean-Daniel Cryans (JIRA)" <ji...@apache.org> on 2011/03/23 01:26:05 UTC

[jira] [Updated] (HBASE-3640) [replication] Transferring queues shouldn't be done inline with RS startup

     [ https://issues.apache.org/jira/browse/HBASE-3640?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Jean-Daniel Cryans updated HBASE-3640:
--------------------------------------

    Attachment: HBASE-3640.patch

This patch adds a simple TPE that runs the failover in the background. I wanted to use the RS's ExecutorService but pulling it down there is quite messy.

> [replication] Transferring queues shouldn't be done inline with RS startup
> --------------------------------------------------------------------------
>
>                 Key: HBASE-3640
>                 URL: https://issues.apache.org/jira/browse/HBASE-3640
>             Project: HBase
>          Issue Type: Improvement
>    Affects Versions: 0.90.1
>            Reporter: Jean-Daniel Cryans
>            Assignee: Jean-Daniel Cryans
>             Fix For: 0.90.2
>
>         Attachments: HBASE-3640.patch
>
>
> Currently ReplicationSourceManager transfers queues of dead region servers in init(). I saw a handful of situations where it had a bunch of znodes to transfer and it delayed the whole region server startup by a few minutes. This instead should be done later, possibly in a Chore.

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira