[Spike] Build missing dashboards to validate pipeline duration proposals
Context
We proposed an approach to reduce pipeline duration in https://gitlab.com/gitlab-com/gitlab-OKRs/-/work_items/5806#note_1770693812, and we'd like to validate that this approach makes sense by looking at existing data.
Goals
- Gather the data we need to validate the proposed approach would help reduce pipeline duration.
- Have correct data for the KR and beyond. We recently noticed that we had issues in our charts, causing us to optimize parts of the pipelines that were not necessarily crucial to improve our metrics.
What data do we need?
From https://gitlab.com/gitlab-com/gitlab-OKRs/-/work_items/5806#note_1770693812 (Next up
section):
- Predictive pipelines
- Transition success rate (like this one, but this one has false positives due to child pipelines for predictive tests)
- Pipeline duration for predictive pipelines
- Percentage of predictive versus full pipelines
- Pipelines duration
- Daily/Weekly/Monthly data (table with pipelines URLs AND bar chart with buckets for pipeline duration on the X axis - like this one, but with the percentiles marked on the X axis, and accurate data)
- Split by pipeline types (predictive, docs, code, E2E, review-apps, …)
- Split MRs pipelines (predictive/full), scheduled master pipelines, non-scheduled master pipelines
Edited by David Dieulivol