blue-green-deployment
Teaches blue-green deployments: dual environments, traffic cutover, validation, rollback, and cloud/K8s implementations. Use when designing or operating blue-green release strategies.
SKILL.mdskills/extras/blue-green-deployment/SKILL.md
GitHubBlue-Green Deployment
1. Overview
What it is: Run two production environments (Blue = current, Green = new). Deploy and test Green, then switch traffic atomically (or near-atomically) from Blue to Green.
Why it exists: Minimize downtime and enable instant rollback by switching traffic back.
When to use: Need fast rollback; dual capacity affordable; DB schema compatible with both during cutover.
When NOT to: Cost-prohibitive double capacity; long-lived incompatible dual writes without migration plan — consider canary or expand-contract first.
2. Concepts
- Blue / Green slots — identical stacks different versions
- Router / LB switch — weighted or 100% cutover
- Smoke / synthetic tests on idle color before switch
- Sticky sessions — drain before switch if stateful
- Shared services — DB, queues must support both versions briefly
- Related: rolling, canary, recreate, A/B, shadow
3. Architecture
graph LR
Users --> LB[Load Balancer]
LB -->|active| Blue[Blue v1]
LB -.->|idle| Green[Green v2]
Blue --> DB[(Shared DB)]
Green --> DB
After cutover:
graph LR
Users --> LB
LB -->|active| Green[Green v2]
LB -.->|standby| Blue[Blue v1]
4. Production Best Practices
- Automate smoke tests against Green before flip
- Keep Blue warm for rapid rollback window (N hours)
- Feature flags for risky behavior beyond binary switch
- Ensure migrations are backward/forward compatible
- Document who can flip traffic and under what criteria
- Monitor error rate/latency during and after switch
5. Common Mistakes
| Mistake | Symptoms | Fix |
|---|---|---|
| Breaking schema change first | Green fails or Blue breaks after | Expand-contract migrations |
| No smoke on Green | Bad release at 100% | Automated pre-flight |
| Instant destroy Blue | Can't rollback | Retain window |
| Session affinity ignored | User errors | Drain connections |
6. Debugging Guide
- Confirm which color serves (
X-Envheader or target group) - Compare target health on both pools
- Diff configs/secrets between colors
- Rollback = point LB to previous color; verify metrics
7. Security
- Identical security controls both colors
- Don't leave old color publicly routable with outdated vulns longer than policy
- Audit trail of traffic switches
8. Performance
- Pre-warm Green (JVM/caches/CDN)
- Capacity equal to Blue before cutover under peak
- DNS TTL considerations if switching at DNS layer (prefer LB weights)
9. Real Production Example
ECS two target groups; CodeDeploy blue/green; hooks run integration tests; flip; CloudWatch alarms auto-rollback to Blue within 15 minutes if 5xx spike.
10. Interview Questions
Beginner: What problem does blue-green solve?
Intermediate: How do DB migrations work with blue-green?
Senior: Blue-green across regions with data replication lag.
11. Checklist
- Dual environment ready + smoke tests
- Compatible migrations
- Rollback window defined
- Alarms + auto/manual rollback
- Runbook for sticky sessions/drain
12. Cheat Sheet
| Platform | Mechanism |
|---|---|
| AWS ALB | Two target groups + weighted forward |
| K8s | Two Deployments + Service selector flip / Ingress canary weights |
| DNS | Weighted records (slower due to TTL) |
Related Skills
Author & maintainer: Shanmukha Kumar Karra
Created and maintained as part of AgentOps Kit.