[
https://issues.apache.org/jira/browse/YUNIKORN-2323?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17805632#comment-17805632
]
Peter Bacsko commented on YUNIKORN-2323:
----------------------------------------
My thoughts:
{quote}ResumeApplication is fired when all ph's are timed out. Do we need to
inform the users about this event as they may not clue any about this
significant change?{quote}
I think it's difficult to efficiently communicate with the user because so many
things are happening. Eg. the status of n placeholders, they time out,
scheduling continues with normal allocation, etc. At a minimum, we should emit
an easy to understand message to the console about what exactly happened and
how the scheduling will proceed. We have more opportunities on the UI:
* Current status of the placeholders: allocation ongoing, replaced, killed, etc
* Current scheduling style for asks: replacing, normal (ie. fallback from soft)
* Problems encountered: placeholder resources are smaller, so placeholder pod
got killed
This can be shown on the REST API.
{quote}When Gang app ph's are in progress (and allocated), when there is
request for real asks and there is resource crunch, do we need to trigger auto
scaling?{quote}
As we discussed on Slack, I have a PR which optimizes autoscaling:
https://github.com/apache/yunikorn-core/pull/745.
However, most likely this is incomplete. We should ignore asks that are about
to be replaced.
Therefore, this condition
{noformat}
if !request.HasTriggeredScaleUp() && request.requiredNode == common.Empty {
...
}
{noformat}
becomes
{noformat}
if !request.HasTriggeredScaleUp() && request.requiredNode == common.Empty &&
!sa.canReplace(request) {
...
}
{noformat}
> Gang scheduling user experience issues
> --------------------------------------
>
> Key: YUNIKORN-2323
> URL: https://issues.apache.org/jira/browse/YUNIKORN-2323
> Project: Apache YuniKorn
> Issue Type: Bug
> Components: core - scheduler
> Affects Versions: 1.4.0
> Reporter: Manikandan R
> Assignee: Manikandan R
> Priority: Major
>
> In case of any issues, users are finding it bit difficult to understand what
> is going on with the gang app.
> Issue 1:
> "driver pod is getting struck"
> At times, when driver pod is not able to run successfully for some reasons,
> users are getting the perspective that pod is getting struck and app is
> hanged, not moving further. Users are waiting for some time and don't
> understand the clear picture. How do we close the gap quickly and communicate
> accordingly through events?
> Issue 2:
> ResumeApplication is fired when all ph's are timed out. Do we need to inform
> the users about this event as they may not clue any about this significant
> change?
> Issue 3:
> When Gang app ph's are in progress (and allocated), when there is request for
> real asks and there is resource crunch, do we need to trigger auto scaling?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
