10 Bonus: sharding

Goal

Understand how to split a suite across shards locally and in CI. Wiring sharding into every workflow is optional.

Read first

• Sharding
• Parallelism

Local shards

npx playwright test --shard=1/4 --project=chromium
npx playwright test --shard=2/4 --project=chromium

Run different shards in different terminals for wall-clock speedup.

This repo sets workers: 1 because the mock API list store is process-global. Sharding across machines or jobs still helps CI. Raising in-process workers needs fixture-owned data first (see Testing guide).

CI sketch (GitHub Actions)

strategy:
  fail-fast: false
  matrix:
    shardIndex: [1, 2, 3, 4]
    shardTotal: [4]
steps:
  - run: npx playwright test --shard=${{ matrix.shardIndex }}/${{ matrix.shardTotal }}

Upload each shard's report or blob as an artifact. Merge blob reports if you use the blob reporter (enabled on CI in this repo).

Done when

npx playwright test --shard=1/4 --project=chromium
npx playwright test --shard=2/4 --project=chromium

both complete on a clone, and you can explain why this repo keeps workers: 1.

Key takeaways

• You can run --shard=N/M locally.
• Sharding is not the same as raising workers without isolation.
• You can sketch a matrix job for CI.

Back to the course home.