Is it possible to trigger a restart of crio to apply new runtime configurations without requiring a node reboot? e.g. through SIGHUP or by watching for changes at /etc/crio/crio.d.

Motivation

When configuring additional runtimes (e.g. the NVIDIA Container Runtime or Kata), we want to make modifications to the crio.conf file (or a file in /etc/crio/crio.d). For example adding:

$ cat crio.conf

[crio]

  [crio.runtime]
    default_runtime = "nvidia"

    [crio.runtime.runtimes]

      [crio.runtime.runtimes.nvidia]
        runtime_path = "/usr/bin/nvidia-container-runtime"
        runtime_type = "oci"

(we don't always set the default runtime)

When using the GPU Operator to deploy the components required to enable access to GPUs in containers, this configuration update is performed by an operand managed by the operator. In the case of docker and containerd the equivalent configuration modifications are made and the docker or containerd daemons are restarted (e.g. using SIGHUP) to ensure that the newly configured runtime is available to run containers -- either using runtime classes, or as the default runtime.

These configuration options are also cleaned up when the operand responsible for applying the configuration is shut down, triggering another reload of the configuration.