load-balancer-config

Overview

Configures load balancers (AWS ALB/NLB, Nginx, HAProxy) for high availability with proper health checks, traffic distribution algorithms (round-robin, least connections, ip-hash), sticky sessions when needed, SSL termination, connection draining, and advanced traffic splitting (blue-green, canary, A/B).

When to Use This Skill

• Running multiple instances of an app behind a load balancer.
• Implementing zero-downtime deploys or canary releases.
• Adding high availability or geographic distribution.

Prerequisites

• Multiple backend instances (or containers/pods) of the same app.
• Health check endpoint on the app (/health or /ready).
• DNS or cloud load balancer service.

Steps

1. Choose LB type:
   • Layer 7 (ALB, Nginx, HAProxy) for HTTP features (path routing, headers, canary).
   • Layer 4 (NLB, HAProxy TCP) for raw TCP/UDP or when you need extreme performance.
2. Health checks:
   • Path, interval, timeout, healthy/unhealthy thresholds.
   • Separate liveness vs readiness if using Kubernetes.
   • Use a dedicated lightweight endpoint that checks critical dependencies (DB, cache) or a simpler one.
3. Algorithms:
   • Round robin (default, even distribution).
   • Least connections (good for long-lived connections).
   • IP hash / source IP (for sticky sessions when state is local).
4. Sticky sessions (when unavoidable):
   • Cookie-based (preferred) or source IP.
   • Document why sticky is needed and plan to remove it.
5. SSL termination:
   • Terminate at LB (recommended) — backends can stay on HTTP.
   • Or pass-through (rare).
6. Connection draining / deregistration:
   • Deregistration delay (AWS) or slow start / drain timeout.
   • Essential for zero-downtime deploys.
7. Traffic splitting:
   • Blue-green: two target groups, switch listener rule or DNS.
   • Canary: weighted target groups (ALB, Nginx, HAProxy, or service mesh).
   • Header or cookie based routing for A/B.
8. Output:
   • Config for the chosen LB (ALB listener + target group Terraform, Nginx upstream + server block, HAProxy frontend/backend).
   • Health check endpoint recommendation for the app.
   • Blue-green and canary deployment scripts or Terraform.
   • Monitoring (LB 5xx, latency, healthy host count).

Examples

AWS ALB Terraform for a blue-green setup, an Nginx config with least-conn and sticky cookie, an HAProxy example with canary weighting, and health check implementation guidance are included.

Edge Cases & Error Handling

• All backends unhealthy: LB should return 503 or a maintenance page. Configure a fallback.
• Session affinity breaking deploys: Use cookie-based affinity with proper drain.
• gRPC / WebSocket: Ensure the LB supports the protocol (ALB supports gRPC, Nginx needs special config).

Verification

1. Deploy multiple backends.
2. Send traffic — it is distributed (check logs or metrics on each backend).
3. Kill one backend — health check removes it, traffic continues on the others.
4. Perform a blue-green or canary deploy — no or minimal user impact.
5. Success: Traffic is balanced, unhealthy instances are automatically removed, and advanced deployment strategies work without downtime.

References

• AWS ALB Documentation
• Nginx Upstream
• HAProxy
• Blue-Green Deployments