Mitsuhiro Shibuya, Mercari
As reliability engineers, many of us would prefer to avoid confronting the network directly. Its behavior is inherently best-effort, lacking the deterministic nature we expect from our application layer. Yet, because we are responsible for the end-to-end reliability of our complex systems, the day inevitably comes when we are forced to face this uncertain domain. This session is for every SRE who decided to dive in.
Instead of just theory, this talk provides a practical playbook for capturing packets in Kubernetes. We'll tackle the key challenges and offer a strategic guide to choosing the right tool for the job—whether it's kubectl debug, ksniff, or modern eBPF-based solutions. You won't leave as a network expert, but as an SRE armed with the toolkit to prove with data. Join us to have the right to say, "See? It's not the network." (or finally prove that it is!)
Mitsuhiro Shibuya is a Tokyo-based Site Reliability Engineer with nearly a decade of experience dedicated to the art of keeping complex systems running. He began his career as a Backend Engineer with a lasting love for Ruby, and keeps that passion alive through modest contributions to open-source projects. This talk was born from his recent experience on a Platform Network team, where he aims to empower fellow engineers with the network troubleshooting skills he was once forced to learn. Cat enthusiast.
