Designing Blue-Green Deployment with Claude Code: Zero Downtime, Instant Rollback

Introduction

Deploy without service interruption and rollback in 5 seconds if issues arise. Blue-Green deployment maintains two environments and instantly switches with the load balancer. Generate designs with Claude Code.

---

CLAUDE.md Blue-Green Rules

## Blue-Green Deployment Rules

### Environment Setup
- Blue: current production (live traffic)
- Green: new version standby
- Both share the same DB/Redis/SQS (Stateless)
- Which color is "live" managed by ALB listener

### Switching Procedure
1. Deploy new version to Green
2. Run smoke tests on Green
3. Switch ALB rule from Blue→Green
4. Monitor metrics for 10 minutes
5. Update Blue to new version (for next Green)

### Rollback Conditions
- Error rate > 1% (5 minutes)
- P99 > 2 seconds
- Health check failure
- Rollback = ALB rule change only (completes in 5 seconds)

---

Generated Blue-Green Deploy

# .github/workflows/blue-green-deploy.yml
jobs:
  blue-green-deploy:
    steps:
      - name: Determine current live color
        id: current
        run: |
          LIVE_COLOR=$(aws elbv2 describe-target-groups \
            --load-balancer-arn ${{ secrets.ALB_ARN }} \
            --query "TargetGroups[?contains(TargetGroupName, 'live')].TargetGroupName" \
            --output text | grep -oE 'blue|green' | head -1)
          echo "live_color=$LIVE_COLOR" >> $GITHUB_OUTPUT
          echo "standby_color=$([ $LIVE_COLOR = blue ] && echo green || echo blue)" >> $GITHUB_OUTPUT

      - name: Deploy to standby + smoke test
        run: |
          aws ecs update-service --cluster production-${{ steps.current.outputs.standby_color }} --service myapp --task-definition myapp-${{ inputs.image_tag }} --force-new-deployment
          aws ecs wait services-stable --cluster production-${{ steps.current.outputs.standby_color }} --services myapp
          STANDBY_URL=${{ secrets.STANDBY_URL }}
          curl -sf "$STANDBY_URL/health" && curl -sf "$STANDBY_URL/ready"

      - name: Switch ALB traffic
        id: switch
        run: |
          STANDBY_TG_ARN=$(aws elbv2 describe-target-groups --names "myapp-${{ steps.current.outputs.standby_color }}" --query "TargetGroups[0].TargetGroupArn" --output text)
          aws elbv2 modify-listener --listener-arn ${{ secrets.LISTENER_ARN }} --default-actions "Type=forward,TargetGroupArn=$STANDBY_TG_ARN"
          echo "Traffic switched to ${{ steps.current.outputs.standby_color }}"

      - name: Monitor 10 minutes
        run: |
          for i in {1..20}; do
            sleep 30
            ERROR_RATE=$(aws cloudwatch get-metric-statistics --namespace AWS/ApplicationELB --metric-name HTTPCode_Target_5XX_Count --start-time $(date -u -d '5 minutes ago' +%Y-%m-%dT%H:%M:%SZ) --end-time $(date -u +%Y-%m-%dT%H:%M:%SZ) --period 300 --statistics Sum --query "Datapoints[0].Sum" --output text 2>/dev/null || echo "0")
            if (( $(echo "$ERROR_RATE > 50" | bc -l) )); then exit 1; fi
          done

      - name: Rollback on failure
        if: failure() && steps.switch.outcome == 'success'
        run: |
          LIVE_TG_ARN=$(aws elbv2 describe-target-groups --names "myapp-${{ steps.current.outputs.live_color }}" --query "TargetGroups[0].TargetGroupArn" --output text)
          aws elbv2 modify-listener --listener-arn ${{ secrets.LISTENER_ARN }} --default-actions "Type=forward,TargetGroupArn=$LIVE_TG_ARN"
          echo "Rolled back to ${{ steps.current.outputs.live_color }}"

---

Summary

Design Blue-Green Deployment with Claude Code:

1. CLAUDE.md — environment definitions, switching procedure, rollback conditions
2. Smoke test standby environment before switching ALB traffic
3. ALB rule change only for rollback (5 seconds, no ECS redeployment)
4. DB schema changes in 3 phases for backward compatibility (both versions running)

---

Review blue-green deployment designs with **Code Review Pack (¥980)* using /code-review at prompt-works.jp*

myouga (@myougatheaxo) — Axolotl VTuber.