[jira] [Commented] (SOLR-16793) Overseer queue length metrics report the same value for all queues

["Paul McArthur (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-16793?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17720667#comment-17720667 ] 

Paul McArthur commented on SOLR-16793:
--------------------------------------

I think there are a couple of options to fix this:
 # Use a different instance of Stats for each ZkDistributedQueue
 # Change the Stats class so that it can keep track of the length of multiple queues (by queue name)

Option 1 initially seems like the better choice because it segregates the metrics for each queue into a different Stats instance. This solution causes problems for the OVERSEERSTATUS API though, because the implementation of that assumes that it can get all of the metrics from a single instance that it accesses from the OCMH, and there would be data missing from the response.

 

Looking for any thoughts on the right way to go here.

[~ab@getopt.org] It was suggested you might have some opinion. Thanks.

 

 

 

 

> Overseer queue length metrics report the same value for all queues
> ------------------------------------------------------------------
>
>                 Key: SOLR-16793
>                 URL: https://issues.apache.org/jira/browse/SOLR-16793
>             Project: Solr
>          Issue Type: Bug
>      Security Level: Public(Default Security Level. Issues are Public) 
>          Components: metrics
>    Affects Versions: 8.6
>            Reporter: Paul McArthur
>            Priority: Minor
>
> We observe that the reported queue length metrics for the Overseer's state update and collection work queues are always identical.
> I believe the cause is a shared instance of the Stats class that the Overseer injects into the ZkDistributedQueue instances for each of it's queues. The Stats class is only able to keep track of the length of one queue however, and this value would be clobbered whenever any of the ZkDistributedQueue instances updates it.
>  



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@solr.apache.org
For additional commands, e-mail: issues-help@solr.apache.org