Can't reach other node services

[wardensfx]

Hi,

I'm having some troubles with resolving dns on my pods that are not on the FIRST node
As a workaround I could use the host's /etc/resolv.conf with adding dnsPolicy: Default on my deployments.yaml
But for standard helm installs it's very uncomfortable (editing the yaml after the installation...)

I investigated a bit, and I think the problem comes from the subnet where the service is loaded (kube-dns on my case, but I would have the same problem for every services I will start)

Each node has the subnet configured on the interface cni0 like 10.42.x.1/24
And with this config, pods can't ping each other if they are on different nodes bc they have a different subnet
Is it normal ?
I guess that's why services are for, but I have a similar issue with my dns that can resolv names on pods that are on the node 1 but, not with the pods en nodes 2,3,4,5

On the otherside service kube-dns is on 10.43. so that's not even on any of the cni0 subnets, so I would expect it to be either all OK or all KO
So maybe it's not about subnets, but I don't know, I ran ouf of ideas

Any idea ?

Thanks

[wardensfx]

It looks it was only the port 53 that wasn't opened to other nodes