将controller-manager的配置 --allocate-node-cidrs 改为 false，错入日志就没有了。
https://blog.kelu.org/tech/2020/05/27/kubernetes-controller-manager-error-cidrnotavailable.html
After reaching out to the Calico community, they quickly pointed out the root cause: the podCIDR allocation was still being handled by the Kubernetes controller manager. When using Calico IPAM, the controller manager should not assign IP addresses to Pods.
https://github.com/k3s-io/k3s/issues/697#issuecomment-576463182