Adding activeDeadlineSeconds to build and pod spec

[daraghlowe]

What are we doing?

• Add a timeout on builds so they won't run infinitely
• Fixes Set a timeout on the builds  #866

[sambhav]

I believe we need both active deadline seconds and ttlSecondsAfterFinished. The former stops builds that are long running and the latter cleans up pods in terminal state.

We would also need to expose these fields in the image spec. We would also need to document what happens with the build controller if we introduce active deadline seconds and the pod is terminated before completion.

Lastly we also need to expose these fields in the image spec.

[daraghlowe]

Thanks for the feedback @samj1912

Unfortunately ttlSecondsAfterFinished only works with jobs.

https://kubernetes.io/docs/concepts/workloads/controllers/ttlafterfinished/

I tested it out anyway to be sure and got this error:

ValidationError(Pod.spec): unknown field "ttlSecondsAfterFinished" in io.k8s.api.core.v1.PodSpec

We're planning on addressing this in our implementation of kpack by creating a new controller to clean up the pods.

We would also need to document what happens with the build controller if we introduce active deadline seconds and the pod is terminated before completion

The container that's running exits with exit code 2 when the deadline seconds are met and the overall pod has a status of Init:Error. I didn't see anything in the logs on the kpack controller when the pod was killed. Is there something in particular I should be looking at here and where does it need to be documented?

Lastly we also need to expose these fields in the image spec.

Updated the image spec also 👍🏻

[matthewmcnew]

Thanks @daraghlowe !