[
https://issues.apache.org/jira/browse/YARN-10796?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17355804#comment-17355804
]
Peter Bacsko commented on YARN-10796:
-------------------------------------
[~gandras] this is a valid concern. Question is, do we accept how it worked
before and say "yeah, that's another way of working"? Are there clusters built
on the fact that a 0% queue cannot scale out properly, despite the max-capacity
setting? Honestly, I don't know. Maybe some people got used to the improper
behavior and expect it to work that way, which does happen in real life.
That said, even a zero capacity queue should be able to occupy the cluster if
nothing else is used, provided max-capacity is set appropriately. So I would
not go for a new property.
> Capacity Scheduler: dynamic queue cannot scale out properly if its capacity
> is 0%
> ---------------------------------------------------------------------------------
>
> Key: YARN-10796
> URL: https://issues.apache.org/jira/browse/YARN-10796
> Project: Hadoop YARN
> Issue Type: Bug
> Components: capacity scheduler, capacityscheduler
> Reporter: Peter Bacsko
> Assignee: Peter Bacsko
> Priority: Major
> Attachments: YARN-10796-001.patch, YARN-10796-002.patch
>
>
> If we have a dynamic queue (AutoCreatedLeafQueue) with capacity = 0%, then it
> cannot properly scale even if it's max-capacity and the parent's max-capacity
> would allow it.
> Example:
> {noformat}
> Cluster Capacity: 16 GB / 16cpu (2 nodes, each with 8 GB / 8 cpu )
> Container allocation size: 1G / 1 vcore
> root.dynamic
> Effective Capacity: <memory: 8192, vCores: 8> ( 50.0%)
> Effective Max Capacity: <memory:16384, vCores:16> (100.0%)
> Template:
> Capacity: 40%
> Max Capacity: 100%
> User Limit Factor: 4
> {noformat}
> leaf-queue-template.capacity = 40%
> leaf-queue-template.maximum-capacity = 100%
> leaf-queue-template.maximum-am-resource-percent = 50%
> leaf-queue-template.minimum-user-limit-percent =100%
> leaf-queue-template.user-limit-factor = 4
> "root.dynamic" has a maximum capacity of 100% and a capacity of 50%.
> Let's assume there are running containers in these dynamic queues (MR sleep
> jobs):
> root.dynamic.user1 = 1 AM + 3 container (capacity = 40%)
> root.dynamic.user2 = 1 AM + 3 container (capacity = 40%)
> root.dynamic.user3 = 1 AM + 15 container (capacity = 0%)
> This scenario will result in an underutilized cluster. There will be approx
> 18% unused capacity. On the other hand, it's still possible to submit a new
> application to root.dynamic.user1 or root.dynamic.user2 and reaching a 100%
> utilization is possible.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
