Observability and Debugging Cloud-Native Apps using Linkerd & Conduit
summary
Demonstrates a service mesh workflow that adds observability without code changes by installing a control plane and injecting sidecar proxies; conduit inject rewrites Kubernetes YAML to add tiny Rust proxies that use about 2 MB per pod and add ~1 ms tail latency. Uses conduit dashboard, conduit stat and conduit tap to find a 500 response on a /fail endpoint between voting and emoji services, narrows failures from cluster-wide to a single caller–callee edge (voting ~50–55% success, web 100%), and confirms the mesh works in a rolling update.
AI-written summary · about
chapters (10)
- 0:00Start
- 0:19Debugging Microservices Is Hard
- 2:16What is a Service Mesh?
- 3:56Example service meshed request
- 5:24How does a service mesh help with debugging?
- 6:14Service mesh debugging
- 7:17Conduit
- 9:12Live debugging of a Kubernetes App
- 26:04Demo Review
- 27:16In conclusion
related talks
41mDebugging an Application with Service Mesh
29mDebugging an app with a service mesh
32mBeyond Security – Using Linkerd to Get Metrics
20mKeynote: Service Meshes and Observability
20mOverview and State of Linkerd
39mSeamless multi-cluster Communication and Observability with Linkerd
05mLightning Talk: Locating and Debugging Failures with Linkerd...
53mDo you need a Service Mesh?