Runtime Endpoint configuration consolidation discussion.

[mikebrow]

Currently kubernetes supports RUNTIME=remote CONTAINER_RUNTIME_ENDPOINT=/var/run/cri-containerd.sock, and crictl supports 3 ways to set it.. that are different:

cli option or env variable:

   --runtime-endpoint value, -r value  Endpoint of CRI container runtime service (default: "/var/run/dockershim.sock") [$CRI_RUNTIME_ENDPOINT]

or:

  --config value, -c value      Location of the client config file (default: "/etc/crictl.yaml") [$CRI_CONFIG_FILE]

where the yaml includes: runtime-endpoint: /var/run/cri-containerd.sock

Further there is an optional

 --image-endpoint value, -i value    Endpoint of CRI image manager service [$CRI_IMAGE_ENDPOINT]

Should we do some consolidation here? Do we need both CONTAINER_RUNTIME_ENDPOINT and CRI_RUNTIME_ENDPOINT?

Does anyone foresee using the image endpoint option in the near future?

If not is CRI_RUNTIME_ENDPOINT not repetitive? Translates to Container Runtime Interface Runtime Endpoint.

Should runtime-endpoint be container-runtime-endpoint or cri-runtime-endpoint?

[mikebrow]

@Random-Liu @mrunalp thoughts?

[feiskyer]

@mikebrow I'm ok with crictl using same env name CONTAINER_RUNTIME_ENDPOINT and CONTAINER_IMAGE_ENDPOINT.

Should runtime-endpoint be container-runtime-endpoint or cri-runtime-endpoint?

I think all is ok and clear. critl doesn't support other endpoints besides CRI. And runtime-endpoint is short and could be typed easily than others.