LGTM with a small thing: can we please make sure there's a log on the daemon-side when we hit a timeout while trying to ping? Thanks!

Something would already get logged

DEBU[0017] Error getting v2 registry: Get https://registry-1.docker.io/v2/: net/http: request canceled while waiting for connection

also seen this

DEBU[0008] Error getting v2 registry: Get https://registry-1.docker.io/v2/: read tcp 52.4.239.183:443: use of closed network connection

@dmcgowan it looks like that's debug level, isn't it? I think what @icecrime is asking is for an error at the error level that say something like request timed out...

Since we are now warning when using a v1 registry it would make sense to elevate this message from Debug to Warn. Checking whether the request failed because it was a timeout would require a larger code change since the error type does not indicate whether a timeout occurred. The hidden http.httpError type does have a Timeout() bool function which can be tested for, however in both the cases I mentioned above the error is actually a url.Error with no way of extracting whether it occurred as a result of a timeout. The correct way to do this is through supporting the CancelRequest method as suggested by http.Client.

I'm okay with this warning.

LGTM