Networking

Nginx Upstream Health Checks

Design upstream blocks that fail visibly, tune probes honestly, and document behavior under partial outages.

Blended evenings 3 weeks · 17 lab hours Intermediate

670,000 KRW informational until enrollment confirms

Reverse proxy diagram sketched on grid paper

Program narrative

Labs bounce traffic between healthy and poisoned upstreams while you tune intervals that do not lie. We rehearse the moment when the load balancer and nginx disagree.

What the labs include

  • Active vs passive health checks tradeoffs
  • max_fails and fail_timeout math on paper first
  • Sticky session hazards with honest warnings
  • Slow start patterns for cold JVMs
  • Access log fields that prove behavior
  • Canary header routing exercise
  • Post-change verification script

Outcomes you can show a lead

  • Tune probes that catch real failures, not trivia
  • Explain why sticky sessions complicate healing
  • Capture logs that settle arguments quickly

Straight FAQ

Envoy or HAProxy?

Compared in discussion; nginx remains the hands-on surface.

TLS termination?

Covered at a config level; certificate procurement is yours.

Limitations?

gRPC specifics are not exercised in labs.

Recent participant notes

“Canary header exercise mirrored our blue/green debates. Still sorting slow start for JVMs, but probes finally match reality.”

Client in manufacturing