jnummelin
commented
8 months ago The procedure should be something like:
- Cordon the node
- drain the node
- stop k0s on the node; if using systemd enabled OS:
systemctl stop koscontroller - Edit the systemd unit at
/etc/systemd/system/k0scontroller.service: remove the--enable-workerand other worker related opts systemctl start k0scontroller- Remove the related node object in k8s api
Is your feature request related to a problem? Please describe.
I have a controller+worker node that is overwhelmed and want it to be served as controller only. Right now, I have to risk destroying the cluster again for that to work out.
Describe the solution you would like
Give us a more comprehensive procedure about how to swap roles
Describe alternatives you've considered
Just keep it as is, but forcefully taint that specific node with NoExecute
Additional context
No response