set `--allocate-node-cidrs=false` on the Kubernetes controller manager when using Calico IPAM

[performantdata]

Calico IPAM doesn't use the .spec.podCIDR or .spec.podCIDRs on the Nodes, so there's no need for letting Kubernetes allocate CIDRs for Nodes. Add instructions to tell Calico IPAM users to set --allocate-node-cidrs=false on the Kubernetes controller manager when initializing a cluster. This will suppress Kubernetes' CIDR allocations and leave that job to the CNI IPAM plugin, where it belongs.

It should prevent messages like these from appearing on the Nodes when the Kubernetes controller manager options are misconfigured:

Events:
  Type    Reason            Age                    From           Message
  ----    ------            ----                   ----           -------
  Normal  CIDRNotAvailable  5m9s (x1134 over 13h)  cidrAllocator  Node worker-0 status is now: CIDRNotAvailable

There's talk among the Kubernetes team of deprecating podCIDR and podCIDRs anyway.

(The above contains speculation. Double check.)

see also: Slack discussion

[performantdata]

Normal  CIDRNotAvailable  5m9s (x1134 over 13h)  cidrAllocator  Node worker-0 status is now: CIDRNotAvailable

On my cluster, adding the --allocate-node-cidrs=false option using the "Reconfiguring a kubeadm cluster" instructions stopped these events.

[ctauchen]

Hi @performantdata, thanks for reporting this! I've created an internal ticket for this one and placed it in the backlog. I'll update here when it is added to an upcoming sprint.

If you'd like to see this done on a more predictable timeline, feel free to open a PR.

[kfox1111]

I hit this same problem when adding some new nodes and the k8s cidr was bigger then the calico one. setting --allocate-node-cidrs=false did resolve it.