Multiple cri sockets found when Docker + crio or Docker + containerd installed

[ericwadams]

This might be more of a feature enhancement than a bug because you have to have installed containers-basic already or run the setup_system.sh script with both crio and containerd as CLRK8S_RUNNER. It seems that you could pick up that variable and on the kubeadm init add in the --cri-socket=crio parameter. I saw there was some related discussion on this at #95 to remove this. Finding a fix took me a bit of time to do so I'd consider adding the parameters based on the CLRK8S_RUNNER variable.

To get around this I edited kubeadm.yaml and manually added in the cri-socket

apiVersion: kubeadm.k8s.io/v1beta1
kind: InitConfiguration
nodeRegistration:
  criSocket: "/var/run/containerd/containerd.sock"
---
apiVersion: kubelet.config.k8s.io/v1beta1
...

Edit: This issue also happens when you try to run ./reset_stack.sh In that case it did not pick up the kubeadm.yaml file setting and I had to manually hack the script to add in --cri-socket="/run/containerd/containerd.sock" to reset it.

[grahamwhaley]

The kata install uses kata-deploy I believe. There have been a few changes there recently-ish on how the config files (in the cluster at least) are edited and updated. I think those changes might come into kata v1.10.
I'm not sure that will fix all/any of the issues you are seeing though. /cc @egernst