hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Virag Kothari (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-11610) Enhance remote meta updates
Date Tue, 19 Aug 2014 22:57:18 GMT

    [ https://issues.apache.org/jira/browse/HBASE-11610?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14103000#comment-14103000

Virag Kothari commented on HBASE-11610:

Thanks for the review.
Based on general feedback here, I will remove the ThreadLocalHtable. Creating a htable for
each put might produce more garbage, but guess that should be fine given that htable is lightweight.
We can still have multiple HConnections to have a higher write throughput, but can get a HTable
instance for a connection picked randomly.  I hope this works for you guys? 
Will do a performance run to check the memory utilization and see if there is any performance
degradation with this approach.

bq. I'm not a big fan of the ThreadLocal<HTableInterface> threadLocalHTable. It's often
difficult to maintain and test.

Agree its difficult to debug threadlocal issues. 
As you said that HConnection.processBatchCallback() is deprecated, so am not very inclined
to use that. Also, this method seems to be more helpful if there is a batch of Puts.

Will address Stack's comments of changing the config name, separating the connection setup
part in a different utility class and adding more docs/comments.

bq. i.e. when would hConnectionPool be null?

This is only during test cases when meta remains no longer on master. This scenario is explained
in the comment few lines above.



> Enhance remote meta updates
> ---------------------------
>                 Key: HBASE-11610
>                 URL: https://issues.apache.org/jira/browse/HBASE-11610
>             Project: HBase
>          Issue Type: Sub-task
>            Reporter: Jimmy Xiang
>            Assignee: Virag Kothari
>         Attachments: HBASE-11610.patch
> Currently, if the meta region is on a regionserver instead of the master, meta update
is synchronized on one HTable instance. We should be able to do better.

This message was sent by Atlassian JIRA

View raw message