What Happens When Service Mesh Maintainers Get a Taste of Their Own Mesh

Explore the challenges and insights gained when service mesh creators become users of their own technology in this conference talk. Discover how the Linkerd maintainers transitioned from developers to on-call operators for their own service mesh. Learn about complex bugs encountered, diagnosed, and resolved while running Linkerd in production. Gain valuable insights into how this shift in perspective influenced product development and improved the overall service mesh experience. Delve into specific cases such as investigating high latency, debugging slow requests, leveraging distributed tracing, and addressing upgrade challenges. Understand the importance of dogfooding and the benefits of sharing cluster resources. Walk away with a deeper understanding of the real-world implications of running a service mesh and the valuable lessons learned by its creators.

Intro
Risha Mars
Carol Scott
Why a Service Mesh?
Investigate high latency
Debugging slow requests
Distributed tracing helps! Sweet!
linkerd check is our friend
Upgrading the proxies is tedious
Can't upgrade linkerd due to a render error
The case of the hanging linkerd check
Dogfooding Dive
Sharing (clusters) is caring
Conclusion