[
https://issues.apache.org/jira/browse/SOLR-14401?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17717383#comment-17717383
]
Patson Luk commented on SOLR-14401:
-----------------------------------
[~dsmiley] Thank you for the quick reply and explanation!
In our setup, we mainly have 2 kinds of query:
1. direct query to a core, distrib = false, isShard unset
2. query to a collection, which would trigger sub-requests
in 8.x, case 1 would increment ".local" metrics, while case 2 would increment
".distrib" + ".local" for the sub-requests spawned
in 9.x, case 1 would increment the "plain" metrics, while case 2 would
increment the "plain" + "[shard]" for the sub-requests spawned
So you are definitely correct that if we change all the calls of case 1 to use
isShard=true, then we will probably have same numbers in 8 vs 9 (as case 1 will
start using "[shard]" as metrics). Though by quickly looking at reference to
ShardParams.IS_SHARD, it might affect some other logic flows too - sorry I'm
not too familiar with such param and its usage.
This is not super critical, even though in 9.x we get different numbers, we can
still kinda tell case 1 and case 2 apart - [shard] metrics is now only
triggered by case 2, but instead of having a number equals to the top query
issued in case 2, we will get number equals to the sub-queries triggered by
such top query.
> "distrib" request handler metrics should only be tracked on pertinent handlers
> ------------------------------------------------------------------------------
>
> Key: SOLR-14401
> URL: https://issues.apache.org/jira/browse/SOLR-14401
> Project: Solr
> Issue Type: Improvement
> Components: metrics
> Reporter: David Smiley
> Assignee: David Smiley
> Priority: Blocker
> Fix For: 9.0
>
> Time Spent: 3h 20m
> Remaining Estimate: 0h
>
> SOLR-13979 in 8.4 added separate request handler metrics for distributed
> requests. However this was done for _all_ request handlers, even though it's
> mainly SearchHandler (and maybe one or two others?) where a distributed
> request is even possible. I refer to this as "metrics pollution" and it's a
> bad thing. It's more weight per handler (latency load & memory), more weight
> for Solr metrics responses, and it's also _suggestive_ that all registered
> handlers can have distributed requests when this is quite false, thus
> confusing people.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
