Why does this exist?

Currently there is an issue when using sidecars (like istio-proxy) with jobs - they don't exit when the job has completed.

This project will monitor those pods and then send a kill signal to the sidecar containers causing them to exit and the job to be marked as succeeded.

How does it work?

It will monitor all pods in the cluster, and terminate the sidecars if the following criteria has been met:

1. The pod was created by a job
2. The pod's non-sidecar containers have exited with exit code 0

Terminate occurs by creating an ephemeral container attached to the sidecar's process namespace and running kill 1.

Limitations

Currently, this implementation cannot handle pods which leverage shareProcessNamespace.

Getting started

1. git clone https://github.com/zachomedia/kubernetes-sidecar-terminator.git
2. kubectl apply -f manifests/