not able to ping or reach cluster ip or cluster/service FQDN from any pod in same namespce

abidahmeds

i dont know if i am only facing this problem , i dont see a option of giving a hostname and using FQDN to reach Pod/cluster. although if i use Pod IP , i can reach the Pod /containers.
i think it is due to kube proxy not working properly in my setup , i dont see any iptables-save output . Does anyonw know if this is a setup issue , i ammusing ubuntu and installing kube-adm as per lab document. my pod-cidr is completely different network from the cluster-cidr ( whihc i think comes default since no step was given in lab document) . since they are different , a Ip route from kube proxy is required but with no tables, obviously service FQDN never works.
who can help please, if you face same problem ??

chrispokorni

Hi,
Can you post a snippet of the command you are running and the error output? It may help in troubleshooting the issue.
Ideally, pods should be able to communicate with each other once calico is installed. Containers can communicate with each other via localhost.
Details of the Kubernetes networking model can be found below:
https://kubernetes.io/docs/concepts/cluster-administration/networking/

Regards,
-Chris

serewicz

Hello,
If you are trying to access a pod from outside the cluster you would need to leverage a service, an ingress controller, or deploy a service mesh like Envoy. Calico handles inside the cluster networking. In lab 3 you used kubectl expose to create a service such that you can access the nginx node from outside the cluster.

If this isn't the issue you're having, it would be helpful to see the command and error as Chis mentioned.

Regards,

abidahmeds

it is more within cluster networking and not external . Pod to Pod communication using cluster IP or FQDN for the service. I saw lot of comments in web regarding setting hairpin- veth in Kubelet config , to check if Kube proxy running properly .. I used standard setup given in lab document Kube adm, Kubectl, Kubelet 1.11.1-00 on ubuntu and applied calico. not sure whihc step i missed , may be i need to resinstall and check again. But i find it strange that reaching other poD using cluster IP/FQDN is not working , where as using PoD IP , it does. so i think it is about routing btw PoDs , possibly even DNS entries ( for FQDN ) .
will check and install again , and add more information here if i face same problem.

serewicz

If I understand you correctly you are able to connect from pod to pod using the IP. You are not able to connect using the DNS name? When you look at all pods kubectl get pods --all-namespaces are the Calico and other pods all running, all containers inside?

What name are you using to test if DNS is working or not? If you can paste some attempts and errors it may be helpful. Looking at output like nslookup for those names could be helpful as well, to see what IP is being referenced for the information.

Regards,