ci(e2e): run the browser suite on 2 workers on the built target too #512

Merged

project_1_bot_05dcee5cd59f879ab35191d664847a9d merged 1 commit from e2e/two-ci-workers into main

2026-10-13 12:44:58 +00:00

project_1_bot_05dcee5cd59f879ab35191d664847a9d commented

2026-10-13 12:16:41 +00:00

(Migrated from gitlab.core-project.space)

Summary

-const CI_WORKERS = match(target)
-  .with("deployed", () => 2)
-  .with("dev", "built", () => 1)
-  .exhaustive();
 …
-  workers: startedByHand ? 2 : CI_WORKERS,
+  workers: 2,

e2e is the longest job of every merge request, and with !510 it becomes the only browser suite there. On one worker, Playwright accounts for 125 s of its 160 s. The runner has 4 CPUs, enough for the two servers plus two workers.

Evidence

Before: pipeline 8, job 83: Running 35 tests using 1 worker → 35 passed (2.2m). Pipeline 9, job 92: 35 passed (2.3m).
After: locally with CI=1 E2E_TARGET=built: Running 35 tests using 2 workers → 35 passed (22.9s). CI numbers from this MR's pipeline to be added once it finishes.

Merge Danger

Door: two-way

The revert is one line.

Blast Radius: flakes

Retries stay at 0, so CPU contention on the shared runner (two jobs at once) would show up as a red e2e, not a silent retry. I'll watch the next few pipelines for flakes.

## Summary ```diff -const CI_WORKERS = match(target) - .with("deployed", () => 2) - .with("dev", "built", () => 1) - .exhaustive(); … - workers: startedByHand ? 2 : CI_WORKERS, + workers: 2, ``` `e2e` is the longest job of every merge request, and with !510 it becomes the only browser suite there. On one worker, Playwright accounts for 125 s of its 160 s. The runner has 4 CPUs, enough for the two servers plus two workers. ## Evidence - **Before:** pipeline 8, job 83: `Running 35 tests using 1 worker` → `35 passed (2.2m)`. Pipeline 9, job 92: `35 passed (2.3m)`. **After:** locally with `CI=1 E2E_TARGET=built`: `Running 35 tests using 2 workers` → `35 passed (22.9s)`. CI numbers from this MR's pipeline to be added once it finishes. ## Merge Danger **Door:** two-way The revert is one line. **Blast Radius:** flakes Retries stay at 0, so CPU contention on the shared runner (two jobs at once) would show up as a red `e2e`, not a silent retry. I'll watch the next few pipelines for flakes.

root commented

2026-10-13 12:44:42 +00:00

(Migrated from gitlab.core-project.space)

added 3 commits

7ecb405e...b88b2982 - 2 commits from branch main
5ab7b62d - ci(e2e): run the browser suite on 2 workers on the built target too

Compare with previous version

added 3 commits <ul><li>7ecb405e...b88b2982 - 2 commits from branch <code>main</code></li><li>5ab7b62d - ci(e2e): run the browser suite on 2 workers on the built target too</li></ul> [Compare with previous version](/amaurycoudr/chess-project/-/merge_requests/12/diffs?diff_id=49&start_sha=7ecb405e736c6eb9e9192fe3df835c14d0ea988d)

root (Migrated from gitlab.core-project.space) merged commit into main

2026-10-13 12:44:58 +00:00

project_1_bot_05dcee5cd59f879ab35191d664847a9d commented

2026-10-13 12:45:42 +00:00

(Migrated from gitlab.core-project.space)

mentioned in merge request !511

project_1_bot_05dcee5cd59f879ab35191d664847a9d commented

2026-10-13 13:19:26 +00:00

(Migrated from gitlab.core-project.space)

After, from main's pipeline 34 (job 316): Running 35 tests using 2 workers → 35 passed (1.6m), i.e. 88 s of Playwright against 125–128 s on 1 worker (pipelines 8 and 9). No flake on this run; retries stay at 0.

**After**, from `main`'s pipeline 34 (job 316): `Running 35 tests using 2 workers` → `35 passed (1.6m)`, i.e. 88 s of Playwright against 125–128 s on 1 worker (pipelines 8 and 9). No flake on this run; retries stay at 0.