[
https://issues.apache.org/jira/browse/HDFS-12461?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16167400#comment-16167400
]
Weiwei Yang commented on HDFS-12461:
------------------------------------
Hi [~anu]
I've tried this scenario in my single node cluster with 3 data dirs, I think
the round robin works as expected.
after a few minutes...
{code}
[wwei@ozone1 hdfs]$ du -sch *
5.3M data1
5.0M data2
4.0M data3
1.1M name
16M total
{code}
a few more minutes...
{code}
[wwei@ozone1 hdfs]$ du -sch *
14M data1
13M data2
13M data3
{code}
finally after 10k keys created,
{code}
[wwei@ozone1 hdfs]$ du -sch *
43M data1
43M data2
42M data3
{code}
I am closing this as Not a Bug now. Feel free to reopen if you have a different
observation or opinion. Thanks.
> Ozone: Ozone data placement is not even
> ---------------------------------------
>
> Key: HDFS-12461
> URL: https://issues.apache.org/jira/browse/HDFS-12461
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: ozone
> Affects Versions: HDFS-7240
> Reporter: Anu Engineer
> Priority: Blocker
> Labels: ozoneMerge
> Fix For: HDFS-7240
>
>
> On a machine with 3 data disks, Ozone keeps on picking the same disk to place
> all containers. Looks like we have a bug in the round robin selection of
> disks.
> Steps to Reproduce:
> 1. Install an Ozone cluster.
> 2. Make sure that datanodes have more than one disk.
> 3. Run corona few times, each run creates more containers.
> 4. Login into the data node.
> 5. Run a command like tree or ls -R /data or independently verify each
> location.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
