[jira] [Commented] (HBASE-25655) Add a new option in PE to indicate the current number of rows in the test table

["Anoop Sam John (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/HBASE-25655?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17298599#comment-17298599 ] 

Anoop Sam John commented on HBASE-25655:
----------------------------------------

To load 100000 rows to table using PE, you would have given a --size arg right?  I assume u have given it as 2 GB 
--size=2
So while doing randomRead PE test also, along with --rows which suggests how many gets every thread to do, u can pass in --size.  Pass this as same value as used in ur write.  This helps to make the range of the random get to be entire data so that it wont hot spot one or few regions

hbase pe --nomapred --oneCon=false  --table=t1 --size=2 --rows=1000 randomRead 10 

> Add a new option in PE to indicate the current number of rows in the test table
> -------------------------------------------------------------------------------
>
>                 Key: HBASE-25655
>                 URL: https://issues.apache.org/jira/browse/HBASE-25655
>             Project: HBase
>          Issue Type: Improvement
>          Components: PE
>            Reporter: Baiqiang Zhao
>            Assignee: Baiqiang Zhao
>            Priority: Major
>
> When we have written 100000 rows in TestTable with 10 preSplits. Then we want to test randomRead with 10 threads, per thread read 1000 rows. But the range of all read keys is in [0, 10000], all in the first region. It may cause hotspot problem, and the result is not accurate.
> This issue add a new option "initRows" to  solve this problem.



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