hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Nick Dimiduk (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (HBASE-10810) LoadTestTool should share the connection and connection pool
Date Sat, 22 Mar 2014 00:10:43 GMT

    [ https://issues.apache.org/jira/browse/HBASE-10810?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13943764#comment-13943764

Nick Dimiduk commented on HBASE-10810:

[~nkeywal] and I came to the same conclusion simultaneously. I think the test should be using
a single pool for all tables it creates, instead of using this HTable constructor with it's
managed pool. In general, I think the HTable constructor is dangerous for anything but a one-off
and should be removed (see HBASE-9117).

> LoadTestTool should share the connection and connection pool
> ------------------------------------------------------------
>                 Key: HBASE-10810
>                 URL: https://issues.apache.org/jira/browse/HBASE-10810
>             Project: HBase
>          Issue Type: Sub-task
>            Reporter: Enis Soztutar
>            Assignee: Enis Soztutar
>             Fix For: hbase-10070
> While running the IT test from HBASE-10572, we've noticed that the number of threads
jumps to 4K's when CM actions are going on. 
> Our [~ndimiduk] summarizes the problem quite good: 
> MultiThreadedReader creates this pool for each HTable:
> {code}
>     ThreadPoolExecutor pool = new ThreadPoolExecutor(1, maxThreads, keepAliveTime, TimeUnit.SECONDS,
>         new SynchronousQueue<Runnable>(), Threads.newDaemonThreadFactory("htable"));
> {code}
> This comes from the HTable creation
> {code}  
> public HTable(Configuration conf, final TableName tableName)
> {code}
> As well the javadoc says Recommended.
> This is wrong.
> In this issue we can change the LTT sub classes to use the shared connection object and
initialize their tables using HConnection.getTable() rather than new HTable(). 
> This is relevant to trunk as well, but there since there is only one outstanding RPC
per thread, it is not such a big problem. 

This message was sent by Atlassian JIRA

View raw message