helm error: Unable to create listener: Error listen tcp6

[yvespp]

On every helm remote command I'm getting the following error:

helm version
E0406 17:26:06.983609   29625 portforward.go:209] Unable to create listener: Error listen tcp6 [::1]:30143: socket: address family not supported by protocol
Client: &version.Version{SemVer:"v2.2.3", GitCommit:"1402a4d6ec9fb349e17b912e32fe259ca21181e3", GitTreeState:"clean"}
Server: &version.Version{SemVer:"v2.2.2", GitCommit:"1b330722aafcb8123114ae51f69d1e884a326f3e", GitTreeState:"clean"}

Despite the error it seams that all commands are working fine though.
Helm is running on RHEL 7.2.

[technosophos]

If HELM_HOST is set to localhost instead of 127.0.0.1, this can happen on systems that try IPv6 connections to localhost first.

[yvespp]

I don’t have HELM_HOST set, I guess it is set to localhost automatically? Maybe this error could be suppressed as everything works fine?

[yvespp]

The error is logged here: https://github.com/kubernetes/kubernetes/blob/47908629a36430e154681b2af73fa91190e0f08d/staging/src/k8s.io/client-go/tools/portforward/portforward.go#L212

I'll try to change it.

[technosophos]

If I had to guess, I'd guess that this is the line that is causing that problem: https://github.com/kubernetes/helm/blob/master/cmd/helm/helm.go#L192

[technosophos]

If you export your HELM_HOST=127.0.0.1, does it work?

[yvespp]

Tiller is unreachable if I set HELM_HOST=127.0.0.1:

$ export HELM_HOST=127.0.0.1
$ helm version
Client: &version.Version{SemVer:"v2.2.2", GitCommit:"1b330722aafcb8123114ae51f69d1e884a326f3e", GitTreeState:"clean"}
Error: cannot connect to Tiller

$ unset HELM_HOST
$ helm version
E0413 08:37:13.656896   18541 portforward.go:209] Unable to create listener: Error listen tcp6 [::1]:22972: socket: address family not supported by protocol
Client: &version.Version{SemVer:"v2.2.2", GitCommit:"1b330722aafcb8123114ae51f69d1e884a326f3e", GitTreeState:"clean"}
Server: &version.Version{SemVer:"v2.2.2", GitCommit:"1b330722aafcb8123114ae51f69d1e884a326f3e", GitTreeState:"clean"}

As far as I understand it, Helm doesn't use the tunnel if HELM_HOST is set and the error doesn't show up.

I think the only way to fix this is to change the logging here (remove the runtime.HandleError): https://github.com/kubernetes/kubernetes/blob/47908629a36430e154681b2af73fa91190e0f08d/staging/src/k8s.io/client-go/tools/portforward/portforward.go#L212

IMHO that line is unnecessary as the error gets returned to the calling function and gets logged there. I'll try to submit a pull request to to kubernetes to remove this.

[yvespp]

This error shows up if ipv6 is disabled, on Linux: sysctl net.ipv6.conf.all.disable_ipv6=1

[kaskavalci]

Thanks @yvespp. For those who is looking for a solution:

sysctl net.ipv6.conf.all.disable_ipv6=0