Page MenuHomePhabricator

Separate mediawiki latency metrics by endpoint
Open, HighPublic

Description

In the context of defining SLOs for the API Gateway, we need to make an educated guess regarding the latency of the REST MediaWiki API.

Right now, the mtail program that analyzes the apache logs doesn't distinguish between calls to various endpoints, making it impossible to extract that information.

We need to update the mtail recipe to support separating metrics by endpoint. In order to do this, we also need to convert the current program to use proper histograms, so that the final code won't be too large.

Event Timeline

Joe created this task.Sep 24 2020, 9:38 AM
Restricted Application added a subscriber: Aklapper. · View Herald TranscriptSep 24 2020, 9:38 AM

Change 629653 had a related patch set uploaded (by Giuseppe Lavagetto; owner: Giuseppe Lavagetto):
[operations/puppet@production] mtail: convert mediawiki to use a real histogram

https://gerrit.wikimedia.org/r/629653

ArielGlenn triaged this task as Medium priority.Sep 28 2020, 9:33 AM
ArielGlenn raised the priority of this task from Medium to High.Sep 28 2020, 9:52 AM

Change 634207 had a related patch set uploaded (by Hnowlan; owner: Hnowlan):
[operations/puppet@production] mtail: create separate metrics histogram for REST API requests

https://gerrit.wikimedia.org/r/634207

Change 629653 merged by Giuseppe Lavagetto:
[operations/puppet@production] mtail: convert mediawiki to use a real histogram

https://gerrit.wikimedia.org/r/629653