Kubernetes 1.18 Caclico CNI CIDRNotAvailable message

Question

I have Kubernets 1.18 cluster with Calico CNI (v3.13.2). I was able to schedule to workload. but in the events I see CIDRNotAvailable message, coming from all nodes in the default name space.

my CIDR range is -cluster-cidr=10.236.0.0/16 in /etc/kubernetes/manifests/kube-controller-manager.yaml file.

kg events -A -w
NAMESPACE     LAST SEEN   TYPE     REASON             OBJECT                                  MESSAGE
default       4m41s       Normal   CIDRNotAvailable   node/kube01                          Node kube01 status is now: CIDRNotAvailable
default       23s         Normal   CIDRNotAvailable   node/kube02                          Node kube02 status is now: CIDRNotAvailable
default       2m56s       Normal   CIDRNotAvailable   node/kube03                          Node kube03 status is now: CIDRNotAvailable
default       4m33s       Normal   CIDRNotAvailable   node/kube04                          Node kube04 status is now: CIDRNotAvailable
default       4m1s        Normal   CIDRNotAvailable   node/kube29                          Node kube29 status is now: CIDRNotAvailable
default       94s         Normal   CIDRNotAvailable   node/kube30                          Node kube30 status is now: CIDRNotAvailable
default       3m12s       Normal   CIDRNotAvailable   node/kube31                          Node kube31 status is now: CIDRNotAvailable

Any idea why it giving this message?

Thanks SR

subnet

folled these steps: https://docs.projectcalico.org/getting-started/kubernetes/quickstart — sfgroups, Apr 13 '20 at 14:22
Have u tried with cidr 192.168.0.0/16? what pod cidr u have in kubeadm init? What is the output of kubectl get nodes -o wide — Arghya Sadhu, Apr 13 '20 at 14:23
this is what my subnet `serviceSubnet: 10.236.0.0/12, podSubnet: 10.236.0.0/16`. I can't change it `192.168` subnet. all nodes in ready status. — sfgroups, Apr 13 '20 at 14:36
both of them in different subnet range, can you look at the image I added in the post — sfgroups, Apr 13 '20 at 15:05
How many nodes do you have in the cluster (with masters included)? — kitt, Apr 13 '20 at 15:42
First of all, your Service subnet definitely overlaps the Pod subnet. But, this may be not the cause of this problem. More details may be found in the logs of `kube-controllermanager`. — kitt, Apr 14 '20 at 05:46
my `serviceSubnet` is causing this issue. I have removed serviceSubnet and let it to kubernets. Now I don't get this error message. — sfgroups, Apr 20 '20 at 17:40

score 2 · Accepted Answer · answered Apr 20 '20 at 18:10

2

I have to remove serviceSubnet from kubeadm configure and use the default one and use this subnet for POD IP podSubnet: 10.201.0.0/16. created cluster with this configurtion, Now this error stop coming, I can see all the node has CIDR subnet assigned.

kubectl get nodes -o jsonpath='{.items[*].spec.podCIDR}'

answered Apr 20 '20 at 18:10

sfgroups

18,151
28
132
204

2

I ran into the same issue after trying to create my cluster using `--pod-network-cidr=10.99.0.0/16` (kubeadm init). If you don't provide a `--service-cidr`, it defaults to the (colliding!) 10.96.0.0/12 (entirely containing my desired pod network). Funny that this is not captured by the (otherwise exhaustive) pre-flight tests of kubeadm init. After all, I first tried to update my serviceSubnet using [this SF answer](https://serverfault.com/questions/952660/change-servicesubnet-in-kubernetes/952677#952677) but ended up tearing everything down and creating my cluster from scratch. – pklaus Jul 17 '20 at 20:55
1

This is still valid today with k8s 1.26. I had to specify both service and pod CIDR in order not to overlap. – Robert Cutajar Jan 29 '23 at 15:23

Kubernetes 1.18 Caclico CNI CIDRNotAvailable message

1 Answers1