Storage driver about ansible-kubernetes-openshift-pi3

[rhuss]

There is no particular reason for using devicemapper AFAIR, but it could be that I had some issues with overlay at the time (tbh, I don't remember ;-).

As you said, to override, just add a

docker:
  storage_driver: overlay2

to config.yml.

[SitoCH]

Perfect, I'll try that. One of the nodes had it's Docker volumes corrupted after an hard shutdown and I remember that I had the same issue on CentOS with the devicemapper driver. After I switched to overlay I never had these problems again so I'll try this also on the Raspberry's cluster.

[SitoCH]

I switched to overlay2 and it seems to be more reliable. The strange thing is that I had to change the option right in the mail.yml role because Ansible seems to ignore the values in the config.yml.
The other thing that I had to modify was the kubeadm command, at the moment it doesn't officially supports overlay2 so I had to force it with --skip-preflight-checks:

command: kubeadm init --config /etc/kubernetes/kubeadm.yml --skip-preflight-checks
command: kubeadm join --token={{ kubeadm_token }} --skip-preflight-checks master:6443 

[SitoCH]

The problem is overlay2, using overlay everything works fine out of the box.

[robertpeteuil]

SitoCH - I had the same problem & resolution.

Based on the Docker website, it appears that Overlay2 is only supported on Debian Stretch (9). This can be seen on the drivers per Distro chart on this page

[rhuss]

@SitoCH @robertpeteuil So is the suggestion to switch to "overlay" in https://github.com/Project31/ansible-kubernetes-openshift-pi3/blob/master/roles/base/tasks/apt.yml#L44 ?

[SitoCH]

Yes, switching to overlay is the best choice. I'm using it successfully and even with a series of hard shutdowns everything kept running fine. Overlay2 isn't supported and when I tried devicemapper I had random corrupted volumes.

[rhuss]

Thanks for the info, just switched to overlay.