[
https://issues.apache.org/jira/browse/SOLR-14665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17160386#comment-17160386
]
Ilan Ginzburg commented on SOLR-14665:
--------------------------------------
There's a workaround: falling back to {{Legacy}} strategy (see
{{Assign.usePolicyFramework()}}).
I assume we'd need in that case some minimal logic for placement, for example
no two replicas of same shard on same node. That's not too hard to implement.
But if we try to support even reasonable use cases it might get challenging
fast. For example Availability Zone affinity placement, maximize free disk etc.
Other option is to fix the Autoscaling assign strategy to make placement
faster, but I think that's a bad idea, unless we decide to fix Autoscaling for
real (which is likely a bad idea as well).
This issue has been here +forever+. The Autoscaling framework is removed in 9.0
(SOLR-14656), so not sure how much more is reasonable to invest in the last
release(s) of 8x on that topic. Maybe we should survey the community first to
understand who's having scaling issues with that today and what they're trying
to do? If we do a short lived fix in 8x (before the real overhaul in 9.0), we
better make it useful and used.
> Collection creation is progressively slower
> -------------------------------------------
>
> Key: SOLR-14665
> URL: https://issues.apache.org/jira/browse/SOLR-14665
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Ishan Chattopadhyaya
> Priority: Blocker
> Fix For: 8.7
>
> Attachments: Screenshot from 2020-07-18 07-49-01.png
>
>
> Plain and simple collection creation (or even shard splits etc.) get
> progressively slower as more and more collections are in the system. The
> culprit is the autoscaling framework (possibly some unnecessary policy
> computation), even when *no autoscaling is being used whatsoever*.
>
> Here is how bad the situation is:
> https://issues.apache.org/jira/browse/SOLR-14656?focusedCommentId=17160311&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17160311
>
> !Screenshot from 2020-07-18 07-49-01.png!
>
> Btw, even when using createnodeset parameter, there is still a gradual
> slowdown (but less steep than the graph here).
>
> This is a matter of grave concern for anyone running a Solr cluster with more
> than a few hundred collections.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
