Feedback to Kubernetes API for sub-process failure

[ianlewis]

We should have a way to know when the sub-process is failing via the Kubernetes API.

Some options might be:

1. Wrapping the health check: Health check is specified to ns-reloader which uses it during it's own health check.
2. Separating ns-reloader and the controller into separate processes, restarting it via a signal.

• This may not be feasible since we want to wrap controllers which are often configured via environment variables and command-line flags.
  • Could maybe be done with wrapper script which loads the shared config.

3. Posting events via the Kubernetes API indicating the status.
4. Returning an error code and terminating the ns-reloader process if the wrapped process starts to crash loop.

[ianlewis]

@Ladicle Option 4 seems like the most straightforward me. I don't really think option 3 is necessary. WDUT?

[Ladicle]

Option 4 sounds good to me. It seems simple and should be enough.