Austin Godber

The behavior you state is fine. The primary use case for the `timeout` supplied through the API is to shorten the timeout period ... e.g. the user doesn't want to...

It's probably a good idea to call out these caveats in the documentation where we talk about using the `timeout` parameter. Specifically call out the potential for data loss on...

I am going to hold on this request for now. It seems that setting `gracePeriodSeconds` on a deployment doesn't propagate to the pods by design: https://github.com/kubernetes/kubernetes/issues/25055 https://github.com/kubernetes/kubernetes/pull/24964 This doesn't mean...

It has occurred to me that at the very least, the job definition should be able to override the `shutdownTimeout`. This could be carefully chosen by the author of the...

One of the problems this issue is meant to address is that when stopping teraslice jobs we often have to wait this full timeout period for reasons that can't be...

We have still seen a 5 minute worker shutdown timeout even with my controlled job shutdown code merged in https://github.com/terascope/teraslice/pull/2074 We're going to have to look at the kafka asset...

I guess we could leave this issue open if we really want jobs to override the shutdown timeout. But that just kind of plasters over whatever the real shutdown problem...

At the very least the Teraslice master could ping the k8s API somehow.

I can't find any evidence that `kubernetes_image` was added for a specific reason. I assume it was me doing the easiest thing to get things working. Instead of me grabbing...

Note that if this option were removed entirely that would prevent "hybrid" teraslice deployments from working. Where the master is outside of kubernetes and the workers and execution controllers run...