Of course we cannot always share details about our work with customers, but nevertheless it is nice to show our achievements and share some solutions.
On dev environments sometimes a simple LXC or VM is powered on with a single partition: Everything is running in the root partition. This is not a wise choice on production systems, but on fire-and-forget machines this is a common practice.
However when deploying a Kubernetes cluster (using Rancher 2), this seems to cause problems.
The error message shows:
Error response from daemon: path /var/lib/rancher is mounted on / but it is not a shared mount]
To overcome this error message, the root partition needs to be mounted using the --make-rshared parameter. This does not remount the partition itself and can be done during runtime:
root@kube3:~# mount --make-rshared /
To overcome a reboot, the command can be added in /etc/rc.local:
root@kube3:~# cat <
mount --make-shared /
root@kube3:~# chmod 755 /etc/rc.local
The cluster provisioning should be able to continue after this adjustment.