[
https://issues.apache.org/jira/browse/HDDS-5713?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17554877#comment-17554877
]
Janus Chow commented on HDDS-5713:
----------------------------------
[~NeilJoshi] Thanks for the advice. I found it very useful.
I've summarized some points as follows:
* The goal of the balance is an average disk usage, which I think the term of
"volume data density" from HDFS can fulfill the goal. As the example shows in
the design doc, all disks are trying to reach an average usage of 70%.
* The phase of Discover, I think in the doc it's the client side operation
which get the node reports from SCM and decide which node needs to do the
balance job.
* The Plan and Execute phase are the detail steps to be implemented in
datanodes, which will decide what container to move and how to move.
I think the overall idea is consistent. What do you think?[~NeilJoshi]
> Add a disk balancer for datanode
> --------------------------------
>
> Key: HDDS-5713
> URL: https://issues.apache.org/jira/browse/HDDS-5713
> Project: Apache Ozone
> Issue Type: New Feature
> Reporter: Jie Yao
> Assignee: Janus Chow
> Priority: Major
>
> since we have container balancer now, disk balancer should be added.
>
> Design doc:
> https://docs.google.com/document/d/11XjOW68BuBDlbLOXR560QtyvI3fWYs3aKzQkxPPMGGY/edit?usp=sharing
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
