Skip to content

Level 3 · Resilience & Service Mesh

Level 3 is about what happens the moment one service in your system gets slow or stops responding — which, given enough time and enough services, will happen. This level builds and deliberately breaks real failure-isolation mechanisms, then shows how a service mesh moves a lot of that logic out of application code entirely.

What was run, and what wasn't

This level runs real services with a real circuit breaker library, real injected latency and real killed processes to trigger failures, and a real lightweight sidecar proxy (Envoy, run standalone) to demonstrate the mesh pattern and mTLS. A full production mesh control plane (Istio/Linkerd's control plane, multi-cluster mesh) was not run as part of this course; where that's the case, it's stated plainly and the underlying mechanism is shown with the standalone proxy instead.

Prerequisites: Levels 1–2.

Modules

  1. Circuit Breakers: Real Failure Isolation: tripping a real breaker against a real failing dependency, and watching it half-open.
  2. Retries, Timeouts & Backoff Done Right: a real retry storm, caused on purpose, and the jittered backoff that fixes it.
  3. Bulkheads & Resource Isolation: isolating one slow dependency's connection pool so it can't starve the rest of the service.
  4. Rate Limiting Between Services: a real token-bucket limiter protecting a downstream service from a real burst.
  5. Service Mesh Fundamentals: The Sidecar Pattern: a real standalone Envoy proxy sitting in front of a service, with traffic routed through it.
  6. Traffic Management: Canary Releases at the Service Level: splitting real traffic between two versions of a service at the proxy layer.
  7. mTLS & Service-to-Service Security: real self-signed certificates, a real handshake, and what breaks when one side doesn't trust the other's cert.
  8. Distributed Tracing Across Services: a real trace ID propagated across three real service calls, visualized end to end.
  9. Consumer-Driven Contract Testing: a contract test that catches a breaking API change before it ships.
  10. Project: Inject Real Failures and Watch Resilience Work: a small chaos script that kills and slows real services while the system keeps serving most requests.

After Level 3

Level 4 steps back from any one service to the whole system: how teams are organized around services, how you migrate an existing monolith incrementally instead of in one risky rewrite, and when microservices are the wrong answer.