This message was deleted.

If I run: kubectl run -it --rm --restart=Never busybox --image=busybox:1.28 -- wget https://dl-cdn.alpinelinux.org/alpine/v3.17/main/aarch64/APKINDEX.tar.gz It downloads successfully. However any other pods that are running, it fails with bad address.

Hmm, there shouldn't be any difference; presumably you've already tried comparing

/etc/resolv.conf

and things like that? Especially if there's multiple entries some how (I wouldn't expect more than one)… Those pods don't have any non-default networking configuration right?

Everything seems fine - hostnames are resolving when using nslookup, so that should rule out the resolv.conf file. No non-default anything that I can see.

Right, that was why I mentioned having multiple nameservers in

resolv.conf

(technically it's undefined which server you use, as it's not supposed to be an ordered list, so it's possible that one thing happened to pick a luckier server). But that doesn't sound relevant here…

Nope. It was using the kube-dns server, but for giggles, I added 8.8.8.8 before it just to see. hostnames were always resolving though. Very odd.

Maybe something is being silly and trying to use IPv6?

I’ve have the same issue with my Mac M1 with wifi only. Lots of weird networking stuff on local cluster and one I built in Parallels. If I had ethernet I’d try that and see if it was any better. The Mac wifi causes me some issues when my networking gets crazy like with k8s and vms.