kube_java_diagnostics

Wrapper around the Python Kubernetes client to get some diagnostic information from JVMs running in pods, specifically thread dumps and class histograms.

Prerequisites

1. A working kubectl configuration
2. python3

Installation

1. Clone the repository
2. pip install -r requirements.txt (in a virtualenv, probably :) )

Usage

Available options:

• --namespace (required): what namespace the pods are in
• --label: If the Java application is in pods that have a label, use this to restrict to which pods the script connects
• --pod: If you want to connect to only a single pod, specify it here
• --container: If the pod contains multiple containers, specify the name of the container running the Java application
• --main: Specify the main class running in the JVM

The script will collect thread dumps and a class histogram from each JVM it connects to, and save them to files named based on the namespace and pod name, like namespace_podname_yyyymmdd-HHMM_threaddump.out and namespace_podname_yyyymmdd-HHMM_histogram.txt