Problem/Motivation

#3549110: [CI] Performance pipeline execution drops warmed caches fixed cache availability, but metrics report indicators are still lost between pipeline runs.

See an upstream bug: https://gitlab.com/gitlab-org/gitlab/-/issues/347537
Related: https://gitlab.com/gitlab-org/gitlab/-/issues/458828

See Slack discussion: https://drupal.slack.com/archives/C079NQPQUEN/p1756449177935989

Proposed resolution

Remaining tasks

User interface changes

Introduced terminology

API changes

Data model changes

Release notes snippet

Issue fork drupal-3558753

Command icon Show commands

Start within a Git clone of the project using the version control instructions.

Or, if you do not have SSH keys set up on git.drupalcode.org:

About issue forks

Comments

mondrake created an issue. See original summary.

alexpott made their first commit to this issue’s fork.

alexpott opened merge request !13875

alexpott’s picture

alexpott
he/they
Primary language English
Location 🇪🇺🌍
commented
Issue summary: View changes
Status: Active » Needs review

I think we just have to run the jobs that generate metrics on the performance pipeline and be done. I've looked for an issue about this specific issue in gitlab's bug tracker but cannot find it - but Chatgpt could - lol - here's the proper upstream issue https://gitlab.com/gitlab-org/gitlab/-/issues/347537

mondrake’s picture

mondrake
Location 🇮🇹
commented
Related issues: +#3521541: [CI] Components tests coverage metrics differ by PHP version

Mind getting in #3521541: [CI] Components tests coverage metrics differ by PHP version first? It’s two months old in the RTBC queue. Would help removing noise here.

mondrake’s picture

mondrake
Location 🇮🇹
commented

In the Slack thread at certain point I think the idea was

  • during 'normal' pipeline runs (commit, MR, daily) let the metrics report file be generated by the corresponding jobs and, besides attaching them as artifacts: reports: metrics:, also save them as artifact files
  • during performance pipelines, fetch the artifacts from above, let the jobs run, but within the job, skip running the actual tool command based on the $PERFORMANCE_TEST variable
  • pass through the fetched metrics report file to artifacts: reports: metrics:

This would shave off execution time.

alexpott’s picture

alexpott
he/they
Primary language English
Location 🇪🇺🌍
commented

@mondrake I think that that complexity will introduce fragility and just is not worth it.

mondrake’s picture

mondrake
Location 🇮🇹
commented
Status: Needs review » Reviewed & tested by the community

OK, let's see if this is fine then.

catch’s picture

catch
he/him
Primary language English
commented
Version: 11.x-dev » 11.3.x-dev
Status: Reviewed & tested by the community » Fixed

We can skip this again when the gitlab bug is fixed (although it's quite old and has been given a low priority). Performance tests run less frequently than they used to after changes to the grafana dashboard, so we can survive some fairly quick jobs running there.

Now that this issue is closed, review the contribution record.

As a contributor, attribute any organization that helped you, or if you volunteered your own time.

Maintainers, credit people who helped resolve this issue.

  • catch committed 88ad1af6 on 11.3.x 
    fix: #3558753 [CI] Performance pipeline execution drops metrics report...

  • catch committed 6625dbc9 on 11.x 
    fix: #3558753 [CI] Performance pipeline execution drops metrics report...

catch closed merge request !13875

mondrake’s picture

mondrake
Location 🇮🇹
commented

Woot! Metrics started working again! Great one

Status: Fixed » Closed (fixed)

Automatically closed - issue fixed for 2 weeks with no activity.