[
https://issues.apache.org/jira/browse/HDFS-17646?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17890912#comment-17890912
]
ASF GitHub Bot commented on HDFS-17646:
---------------------------------------
huangzhaobo99 commented on PR #7120:
URL: https://github.com/apache/hadoop/pull/7120#issuecomment-2422243631
> Good idea. Only one concerns, if give one small limitTopNodes is it
possible to impact DataNode service when trigger balancer. Such as one corner
case, given set limitTopNodes=1 and there are thousands underUtilized nodes,
then the top one datanode will involve transfer flood, right? So IMO we should
give some caution information to end user at least. FYI. Thanks.
Similar to using the '-source' param, top one datanode will involve transfer
flood, this corner case will continue to migrate data from the node until the
conditions for Balance are met.
However, the blocks and speed of each migration are affected by other
params, and the impact on the cluster is controllable.
> Add Option to limit Balancer prefer highly utilized nodes num in each
> iteration
> -------------------------------------------------------------------------------
>
> Key: HDFS-17646
> URL: https://issues.apache.org/jira/browse/HDFS-17646
> Project: Hadoop HDFS
> Issue Type: New Feature
> Reporter: Zhaobo Huang
> Assignee: Zhaobo Huang
> Priority: Major
> Labels: pull-request-available
>
> Limit the number of topN nodes to avoid excessive nodes affecting cluster
> stability.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
