[
https://issues.apache.org/jira/browse/SOLR-14401?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17494892#comment-17494892
]
David Smiley commented on SOLR-14401:
-------------------------------------
In my PR, I did something a bit different than what I said above; feedback
welcome! I placed/named these sharded metrics by appending a “[shard]” to the
request handler name. Thus in addition to "{{QUERY./select.requests}}"
(unchanged) there is now "{{QUERY./select[shard].requests}}". It has a kind of
appeal to me… I wonder if some day request handlers might even be addressable
this way (in the URL) to thus make it more apparent from a logging standpoint
that these requests are special without having to dig into the particular
parameters to see this important fact. If the {{.shard.}} is added as we said,
it would hang below and have the same names that the parent does, which is
adequate but ... I dunno. It means one named thing below ./select. isn't
actually a metric like the others but is another namespace of metrics (for the
shard requests).
Other notes:
* Capture all the same metrics for the request, not just the couple that were
being tracked in this fashion. Thus we now have more metrics and split up
between the requests like error tracking at the coordinator distinct from the
shard.
* moved the shard purpose statistic formerly grouped under "shardRequests" to
be under the shard group, and moved this processing to be in SearchHandler as
it is specific to this class. {{QUERY./select.shardRequests}} ->
{{QUERY./select[shard].purposes}}
And it no longer contains a “total” counter within it because it’s redundant
with "{{QUERY./select[shard].requests}}"
* The overall logic formerly was only activated in SolrCloud mode but I see no
point in that restriction. Users have done distributed search years before
SolrCloud and still do today!
> "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: 10m
> 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
(v8.20.1#820001)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
