docker-env should ignore invalid docker proxy env vars

[lingsamuel]

Steps to reproduce the issue:

1. minikube start --driver=kvm2
2. minikube docker-env

I found sometimes docker-env finished almost instantly, sometimes it takes about 10s+.

[afbjorklund]

You can try troubleshooting with --alsologtostderr, to see where it is spending the time ?

There were a lot of health checks added, both to the cluster and to the docker runtime itself.

[lingsamuel]

Okay I found the reason...

I0108 11:08:47.237526 3197810 docker-env.go:427] Testing Docker connectivity with: /usr/bin/docker version --format={{.Server}}
W0108 11:08:47.276089 3197810 docker-env.go:293] couldn't connect to docker inside minikube.  output: proxy: unknown scheme: socks
 error: exit status 1
W0108 11:08:47.276112 3197810 docker-env.go:187] dockerd is not active will try to reload it...

I have proxy set to socks://127.0.0.1:1080/
golang requires socks5 xref golang/go#16715 and https://github.com/golang/net/blob/6772e930b67bb09bf22262c7378e7d2f67cf59d1/proxy/proxy.go#L91-L99

[lingsamuel]

So that's the reason why my pods always restart 😂 I was thought that's because of my lack of memory

[afbjorklund]

There should be some output, when minikube is doing "extra" stuff (instead of printing docker-env)

I think in docker-machine env completes much faster, but you might end up with bogus variables.

Normally it is fast, though.

Just checks the cluster status and runtime connectivity (docker version).

[lingsamuel]

There should be some output, when minikube is doing "extra" stuff

Couldn't agree more! Sometimes I have to add --alsologtostderr to see why minikube hangs there.