kubelet does not have ClusterDNS IP configured

[vishnumumbai]

Hi,
I am getting the following error while describing the pod.

pod: "nginx-54c98b4f84-42m6f_default(0e15b9f4-6898-4783-a729-24855dbd342a)". kubelet does not have ClusterDNS IP configured and cannot create Pod using "ClusterFirst" policy. Falling back to "Default" policy.

[cdkharris]

I got the same error. In my case, after deleting the pod and forcing it to recreate I found that this error was preceded by a different error:

3s                      Warning   FailedCreatePodSandBox   Pod/nginx-59c77757f7-bvp4b    Failed to create pod sandbox: rpc error: code = Unknown desc = failed to setup network for sandbox "eb6b32c896504fdfc1f4ab813665fa8b107c83cb60624b5670c3f9ba76fe908c": plugin type="bridge" name="bridge" failed (add): invalid CIDR address:
3s (x91705 over 13d)    Warning   MissingClusterDNS        Node/node-1                   kubelet does not have ClusterDNS IP configured and cannot create Pod using "ClusterFirst" policy. Falling back to "Default" policy.

This led me to look at my 10-bridge.yaml configuration file which is created in this step. I noticed that I had not configured the subnet.

In my case, this occured because when I initially ran the commands to create the 10-bridge.yaml and the kubelet-config.yaml configuration files the machines.txt file was not in the same directory. So the subnet was not available to populate the configuration files. I fixed the configuration file, distributed it to the worker nodes as shown, and then restarted the containerd, kubelet, and kube-proxy services on the worker nodes. After that the pod was able to successfully be created.

[osimanov]

I'm getting
kubelet does not have ClusterDNS IP configured and cannot create Pod using "ClusterFirst" policy. Falling back to "Default" policy.
error as well. As far as I understand it means --cluster-dns parameter was not set when running kubelet. From kubelet help:
--cluster-dns strings Comma-separated list of DNS server IP address. This value is used for containers DNS server in case of Pods with "dnsPolicy=ClusterFirst"
And dnsPolicy default value is ClusterFirst as per pod.spec.dnsPolicy spec:
Set DNS policy for the pod. Defaults to "ClusterFirst".
My humble guess is: the author has skipped this parameter because there is no cluster wide dns service installed.

[ermazi]

author just omit cluster dns, so pod launched with resolve file in node.
if u wanna use cluster dns, plz follow bellow steps:

1. add install helm
2. add coredns helm chart repo and install coredns in kube-system
3. get cluster ip from coredns service
4. set --cluster-dns in kubelet component on every node
5. recreate all pods u wanna use cluster dns

btw: fix #877 first