[jira] [Commented] (HBASE-25528) Dedicated merge dispatch threadpool on master

["Bharath Vissapragada (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-25528?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17271675#comment-17271675 ] 

Bharath Vissapragada commented on HBASE-25528:
----------------------------------------------

We were also bottlenecked on heavy meta lookup operations for each candidate plan in the normalizer but I just realized that was already fixed by HBASE-24628 (the change that uses MasterServices directly rather than making an admin RPC). There is an unnecessary meta lookup in the merge code path that was causing severe slow down. I'll submit a separate patch to get rid of that.

> Dedicated merge dispatch threadpool on master
> ---------------------------------------------
>
>                 Key: HBASE-25528
>                 URL: https://issues.apache.org/jira/browse/HBASE-25528
>             Project: HBase
>          Issue Type: Improvement
>          Components: master
>    Affects Versions: 3.0.0-alpha-1, 1.7.0, 2.5.0
>            Reporter: Bharath Vissapragada
>            Assignee: Bharath Vissapragada
>            Priority: Minor
>
> Over the weekend we ran into an issue of running normalizer on a table with ~400k regions and we noticed the master was bottlenecked on a single threaded thread-pool that was meant for all master table operations (opcode: MASTER_TABLE_OPERATIONS). Having a dedicated thread-pool helped us dispatch merges at a faster rate.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)