Build safe, fast rollback mechanisms for production deployments.
Manual Rollback Workflow
# .github/workflows/rollback.yml
name: Rollback
on: workflow_dispatch: inputs: version: description: "Version to rollback to (e.g., v1.2.3 or previous)" required: true type: string environment: description: "Environment to rollback" required: true type: choice options: - staging - production reason: description: "Reason for rollback" required: true type: string
jobs: rollback: runs-on: ubuntu-latest environment: ${{ github.event.inputs.environment }} steps: - uses: actions/checkout@v4 with: ref: ${{ github.event.inputs.version }}
- name: Verify version exists
run: |
if ! git rev-parse ${{ github.event.inputs.version }} >/dev/null 2>&1; then
echo "❌ Version ${{ github.event.inputs.version }} not found"
exit 1
fi
echo "✅ Version ${{ github.event.inputs.version }} exists"
- name: Get current version
id: current
run: |
CURRENT=$(git describe --tags --abbrev=0)
echo "version=$CURRENT" >> $GITHUB_OUTPUT
echo "Current version: $CURRENT"
- name: Confirm rollback
run: |
echo "🔄 Rolling back from ${{ steps.current.outputs.version }} to ${{ github.event.inputs.version }}"
echo "Environment: ${{ github.event.inputs.environment }}"
echo "Reason: ${{ github.event.inputs.reason }}"
- uses: actions/setup-node@v4
with:
node-version: "20"
- run: npm ci
- run: npm run build
- name: Deploy rollback
run: |
./scripts/deploy.sh ${{ github.event.inputs.environment }}
env:
DEPLOY_TOKEN: ${{ secrets.DEPLOY_TOKEN }}
- name: Verify deployment
run: |
./scripts/health-check.sh ${{ github.event.inputs.environment }}
- name: Create incident issue
uses: actions/github-script@v7
with:
script: |
github.rest.issues.create({
owner: context.repo.owner,
repo: context.repo.repo,
title: `Rollback: ${context.payload.inputs.environment} to ${context.payload.inputs.version}`,
body: `## Rollback Details
**Environment:** ${context.payload.inputs.environment}
**From:** ${{ steps.current.outputs.version }}
**To:** ${context.payload.inputs.version}
**Reason:** ${context.payload.inputs.reason}
**Triggered by:** @${context.actor}
**Time:** ${new Date().toISOString()}
## Next Steps
- [ ] Verify rollback successful
- [ ] Investigate root cause
- [ ] Create fix
- [ ] Update postmortem
`,
labels: ['incident', 'rollback']
})
## Automated Rollback on Failure
```yaml
deploy:
runs-on: ubuntu-latest
steps:
- name: Deploy
id: deploy
run: ./scripts/deploy.sh production
continue-on-error: true
- name: Verify deployment
id: verify
if: steps.deploy.outcome == 'success'
run: ./scripts/health-check.sh production
continue-on-error: true
- name: Auto-rollback on failure
if: steps.deploy.outcome == 'failure' || steps.verify.outcome == 'failure'
run: |
echo "⚠️ Deployment failed, initiating automatic rollback"
PREVIOUS_VERSION=$(git describe --tags --abbrev=0 HEAD^)
./scripts/deploy.sh production $PREVIOUS_VERSION
# Verify rollback
if ./scripts/health-check.sh production; then
echo "✅ Rollback successful"
else
echo "❌ Rollback failed - manual intervention required"
exit 1
fi
Kubernetes Rollback
rollback-k8s:
runs-on: ubuntu-latest
steps:
- name: Setup kubectl
uses: azure/setup-kubectl@v3
- name: Configure kubectl
run: |
echo "${{ secrets.KUBECONFIG }}" > kubeconfig
export KUBECONFIG=kubeconfig
- name: Rollback deployment
run: |
kubectl rollout undo deployment/myapp -n production
kubectl rollout status deployment/myapp -n production --timeout=5m
- name: Get rollback revision
run: |
kubectl rollout history deployment/myapp -n production
Docker Image Rollback
- name: Rollback to previous image
run: |
# Get previous image tag
PREVIOUS_TAG=$(docker inspect myapp:latest | jq -r '.[0].ContainerConfig.Labels.previous_tag')
# Retag previous as latest
docker pull myapp:$PREVIOUS_TAG
docker tag myapp:$PREVIOUS_TAG myapp:latest
docker push myapp:latest
# Restart containers
docker-compose pull
docker-compose up -d
Database Migration Rollback
- name: Rollback database migrations
run: |
# Get migration to rollback to
CURRENT=$(npm run migrate:current)
TARGET=${{ github.event.inputs.migration }}
echo "Rolling back from $CURRENT to $TARGET"
npm run migrate:down -- --to=$TARGET
# Verify rollback
AFTER=$(npm run migrate:current)
if [ "$AFTER" != "$TARGET" ]; then
echo "❌ Migration rollback failed"
exit 1
fi
env:
DATABASE_URL: ${{ secrets.DATABASE_URL }}
Rollback Runbook
# Production Rollback Runbook
## When to Rollback
Rollback if:
- Critical bugs affecting >10% of users
- Data integrity issues
- Security vulnerabilities
- Performance degradation >50%
- Error rate >5%
## Before Rollback
1. **Assess impact**: Check monitoring dashboards
2. **Identify version**: Determine last known good version
3. **Notify team**: Post in #incidents Slack channel
4. **Enable maintenance mode** (if possible)
## Rollback Steps
### Automated Rollback (Preferred)
1. Go to Actions → Rollback workflow
2. Select environment (staging/production)
3. Enter target version (e.g., v1.2.3 or "previous")
4. Enter reason for rollback
5. Click "Run workflow"
6. Monitor progress in Actions tab
### Manual Rollback (Emergency)
```bash
# 1. SSH to production server
ssh production
# 2. Check current version
docker ps | grep myapp
# 3. Pull previous version
docker pull myapp:v1.2.3
# 4. Update docker-compose
vim docker-compose.yml
# Change image: myapp:latest to myapp:v1.2.3
# 5. Deploy
docker-compose up -d
# 6. Verify
curl https://api.myapp.com/health
# 7. Check logs
docker logs myapp -f
```
After Rollback
- Verify: Run smoke tests
- Monitor: Watch error rates for 15 minutes
- Notify: Update #incidents with status
- Disable maintenance mode
- Create incident ticket
- Schedule postmortem
Rollback Verification
- Health check returns 200
- Error rate <1%
- Response time p95 <500ms
- Key features working (login, checkout, etc.)
- Database connectivity OK
Communication Template
🔄 ROLLBACK IN PROGRESS
Environment: Production
From: v1.3.0
To: v1.2.3
Reason: Critical bug in checkout flow
Status: In progress
ETA: 5 minutes
Updates: #incidents
Common Issues
Issue: Rollback Fails
Symptom: Deployment doesn't start Fix: Check logs, verify version exists, ensure secrets are valid
Issue: Database Incompatibility
Symptom: App starts but can't read data Fix: May need to rollback migrations first
Issue: Traffic Not Routing
Symptom: Users still see new version Fix: Clear CDN cache, check load balancer config
## Health Check Script
```bash
#!/bin/bash
# scripts/health-check.sh
ENVIRONMENT=$1
BASE_URL="https://${ENVIRONMENT}.myapp.com"
echo "Running health checks for $ENVIRONMENT..."
# API health
if ! curl -f "$BASE_URL/api/health" > /dev/null 2>&1; then
echo "❌ API health check failed"
exit 1
fi
# Database connection
if ! curl -f "$BASE_URL/api/health/db" > /dev/null 2>&1; then
echo "❌ Database health check failed"
exit 1
fi
# Key endpoints
ENDPOINTS=("/api/users" "/api/products" "/api/orders")
for endpoint in "${ENDPOINTS[@]}"; do
if ! curl -f "$BASE_URL$endpoint" > /dev/null 2>&1; then
echo "❌ Endpoint $endpoint health check failed"
exit 1
fi
done
echo "✅ All health checks passed"
exit 0
Best Practices
- Fast rollback: <5 minutes to previous version
- Automated: One-click rollback workflow
- Verified: Health checks after rollback
- Documented: Clear runbook
- Tested: Practice rollbacks regularly
- Monitored: Alert on failures
- Communicated: Notify stakeholders
Output Checklist
- Manual rollback workflow
- Automated rollback on failure
- Platform-specific rollback (K8s/Docker)
- Database rollback procedure
- Rollback runbook documented
- Health check scripts
- Communication templates
- Incident issue automation