[ 
https://issues.apache.org/jira/browse/SOLR-11833?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16439761#comment-16439761
 ] 

Andrzej Bialecki  commented on SOLR-11833:
------------------------------------------

This patch contains the following changes and additions:
 * configurable upper and lower thresholds, as well as actions to be takenĀ on 
each condition
 * separate actions to be taken when node-level rates are above / below 
thresholds
 * configurable metric that determines the "search rate" (by default it's 
{{QUERY./select.requestTimes:1minRate}})
 * configurable limits on the minimum / maximum number of operations to 
request, depending on how much the thresholds are exceeded and on the 
replication factor and the current number of searchable replica types.
 * unit and integration tests

Documentation to follow shortly.

> Allow searchRate trigger to delete replicas
> -------------------------------------------
>
>                 Key: SOLR-11833
>                 URL: https://issues.apache.org/jira/browse/SOLR-11833
>             Project: Solr
>          Issue Type: Sub-task
>      Security Level: Public(Default Security Level. Issues are Public) 
>          Components: AutoScaling
>            Reporter: Andrzej Bialecki 
>            Assignee: Andrzej Bialecki 
>            Priority: Major
>             Fix For: 7.4, master (8.0)
>
>         Attachments: SOLR-11833.patch
>
>
> Currently {{SearchRateTrigger}} generates events when search rate thresholds 
> are exceeded, and {{ComputePlanAction}} computes ADDREPLICA actions in 
> response - adding replicas should allow the search rate to be reduced across 
> the increased number of replicas.
> However, once the peak load period is over the collection is left with too 
> many replicas, which unnecessarily tie cluster resources. 
> {{SearchRateTrigger}} should detect situations like this and generate events 
> that should cause some of these replicas to be deleted.
> {{SearchRateTrigger}} should use hysteresis to avoid thrashing when the rate 
> is close to the threshold.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org

Reply via email to