[
https://issues.apache.org/jira/browse/HDFS-8818?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15995450#comment-15995450
]
Daryn Sharp commented on HDFS-8818:
-----------------------------------
bq. The new design is more flexible than the previous one since we can control
the number of thread per datanode pair.
That sounds good on paper but flexibility does not negate the fact it's proven
not to scale. I'm sure the redesign works great on a couple dozen node
cluster. As illustrated by Kihwal, it limps along on a 280 node cluster
running slower than before and is virtually unusable on multi-thousand node
clusters even with HDFS-11377.
This has to be fixed in a manner that restores previous performance or be
reverted. A jira touting "run faster" can't make the balancer slower and unfit
for production...
> Allow Balancer to run faster
> ----------------------------
>
> Key: HDFS-8818
> URL: https://issues.apache.org/jira/browse/HDFS-8818
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: balancer & mover
> Reporter: Tsz Wo Nicholas Sze
> Assignee: Tsz Wo Nicholas Sze
> Fix For: 2.8.0, 2.7.4, 3.0.0-alpha1
>
> Attachments: bal1.png, bal2.png, h8818_20150723.patch,
> h8818_20150727.patch, HDFS-8818-branch-2.7.00.patch
>
>
> The original design of Balancer is intentionally to make it run slowly so
> that the balancing activities won't affect the normal cluster activities and
> the running jobs.
> There are new use case that cluster admin may choose to balance the cluster
> when the cluster load is low, or in a maintain window. So that we should
> have an option to allow Balancer to run faster.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
