Warn about bad cluster configurations

[twz123]

There have been some reports about high k0s resource consumption recently. Those reports were related to bad cluster configurations. What about adding some sort of warning into k0sctl once it detects such a configuration?

Things to warn about:

• No load balancer (i.e. neither api.externalAddress is set nor network.nodeLocalLoadBalancing is enabled) for clusters with more than one controller
• An even number of controllers when internal etcd is used
• Node-local load balancing is enabled, but conflicts with other cluster settings.

Maybe also worth a warning, albeit there's no user report about it yet:

• More than one controller is configured when using kine+sqlite.

[xinity]

hey there,

I think i typically felt into the first item :(

thanks for the warning,

[twz123]

Added NLLB to the checklist. See k0sproject/k0s#4056.

[jsalgado78]

I agree. I've tried to upgrade my K0s cluster from 1.28.4 to 1.29.2 and kubelet failed because of I use this featureGates deprecated option:

      spec:
        workerProfiles:
          - name: myprofile
            values:
              featureGates: 
                SeccompDefault: true

It would be interesting if k0sctl had warned me about this option before upgrading