GitHub Actions Concurrency Patterns
The Problem: 5 Instances, All Pushing to Main
Running 5 parallel Claude Code instances means bursts of commits to main. Without concurrency configuration, deploy workflows pile up:
- 5 simultaneous deploys competing for resources
- Possible race condition where an older build overwrites a newer one
- Timeouts from resource contention
cancel-in-progress: false is the Right Call for Deploys
# .github/workflows/deploy-prod.yml
concurrency:
group: deploy-prod
cancel-in-progress: false # queue mode — all runs execute in order
The key distinction:
| Use case | cancel-in-progress |
|---|---|
| CI (PR builds) |
true — old PR builds are throwaway |
| Production deploy |
false — every commit should ship |
| Blog publishing |
false — don't skip any post |
| Scheduled tasks |
false — never skip a scheduled run |
Concurrency Group Design
# ❌ Single group for everything — all workflows block each other
concurrency:
group: ci
# ✅ Separate by workflow
concurrency:
group: ${{ github.workflow }}
cancel-in-progress: false
# ✅ Per-PR CI — cancel old builds for the same PR
concurrency:
group: ${{ github.workflow }}-${{ github.ref }}
cancel-in-progress: true
Full Deploy Config for Parallel Pushes
name: Deploy to Production
concurrency:
group: deploy-prod
cancel-in-progress: false
on:
push:
branches: [main]
paths:
- 'lib/**'
- 'web/**'
- 'pubspec.yaml'
paths-ignore:
- 'docs/**'
- '**.md'
jobs:
deploy:
runs-on: ubuntu-latest
timeout-minutes: 30
steps:
- uses: actions/checkout@v4
- run: firebase deploy --only hosting
Always Set timeout-minutes
Without it, a hung job runs for 6 hours and blocks the concurrency queue:
jobs:
build:
timeout-minutes: 20 # Flutter Web build: ~10-15 min
deploy:
timeout-minutes: 10 # Firebase deploy: ~3-5 min
lint:
timeout-minutes: 5 # flutter analyze: ~2 min
Filter Triggers With paths-ignore
on:
push:
paths-ignore:
- 'docs/**'
- '**.md'
- '.claude/**'
- 'memory/**'
Doc-only commits don't need a full deploy. With 5 active instances writing docs frequently, this saves significant CI minutes.
Summary
deploys: cancel-in-progress: false (queue, no skips)
PR checks: cancel-in-progress: true (latest wins)
scheduled tasks: cancel-in-progress: false (no skips)
timeout-minutes: always set on every job
paths-ignore: exclude docs/*.md to reduce unnecessary triggers
Concurrency config is a force multiplier for parallel dev workflows — get it wrong and your CI bills balloon; get it right and 5 instances push in harmony.
Building in public: https://my-web-app-b67f4.web.app/
