Skip to content

Parallelize `rspec foss-impact` by test level based on Knapsack data

Rémy Coutable requested to merge generalize-rspec-pipeline-generation into master

What does this MR do and why?

Prerequisite of https://gitlab.com/gitlab-com/gitlab-OKRs/-/work_items/122796732.

  • This generalize the RSpec pipeline generation script and port it to Ruby, so that later we can reuse the script for https://gitlab.com/gitlab-com/gitlab-OKRs/-/work_items/122796732.
  • This also split the rspec foss-impact jobs by test level and parallelize based on the latest Knaspack report (it calculates the average duration of test files per test level).

You can try runnnig the script locally:

$ curl -o matching_tests-foss.txt https://gitlab-org.gitlab.io/-/gitlab/-/jobs/3794959154/artifacts/rspec/matching_tests-foss.txt

$ curl -o knapsack-report-master.json https://gitlab-org.gitlab.io/gitlab/knapsack/report-master.json

$ scripts/generate_rspec_pipeline.rb -f "matching_tests-foss.txt" -t ".gitlab/ci/rails/rspec-foss-impact.gitlab-ci.yml.erb" -k knapsack-report-master.json
Optimal node count for 0 migration RSpec files is 0.
Optimal node count for 0 background_migration RSpec files is 0.
Optimal node count for 69 unit RSpec files is 2.
Optimal node count for 8 integration RSpec files is 1.
Optimal node count for 10 system RSpec files is 2.

# and then inspect the generated pipeline file:

$ cat .gitlab/ci/rails/rspec-foss-impact.gitlab-ci.yml.erb.yml

Example of a pipeline generated with dynamic parallelization: https://gitlab.com/gitlab-org/gitlab/-/pipelines/781962519

MR acceptance checklist

This checklist encourages us to confirm any changes have been analyzed to reduce risks in quality, performance, reliability, security, and maintainability.

Edited by David Dieulivol

Merge request reports