Profiling an application deployed in a Kubernetes cluster can be a challenge, as profilers often rely on specific protocols that require access to the application's process. With Kubernetes, you need to successfully access the process of a running pod within a sometimes hardened image, start profiling your application, and then retrieve the profile data. Each step can be complicated.

Yahoo's kube-flame tool makes it easy to launch a profiling session and retrieve a FlameGraph-like profile via a kubectl plugin.

In this session, I will show you how to use kube-flame to profile a Java application deployed in a Kubernetes cluster.

свернуть