mirror of
https://github.com/neondatabase/neon.git
synced 2025-12-22 21:59:59 +00:00
e65f0fe874
## Problem We've got several issues with the current `benchmarks` job setup: - `benchmark_durations.json` file (that we generate in runtime to split tests into several jobs[0]) is not consistent between these jobs (and very not consistent with the file if we rerun the job). I.e. test selection for each job can be different, which could end up in missed tests in a test run. - `scripts/benchmark_durations` doesn't fetch all tests from the database (it doesn't expect any extra directories inside `test_runner/performance`) - For some reason, currently split into 4 groups ends up with the 4th group has no tests to run, which fails the job[1] - [0] https://github.com/neondatabase/neon/pull/4683 - [1] https://github.com/neondatabase/neon/issues/6629 ## Summary of changes - Generate `benchmark_durations.json` file once before we start `benchmarks` jobs (this makes it consistent across the jobs) and pass the file content through the GitHub Actions input (this makes it consistent for reruns) - `scripts/benchmark_durations` fix SQL query for getting all required tests - Split benchmarks into 5 jobs instead of 4 jobs.