Docker CLI is trying to use Docker Registry API V1 instead of V2

[Vespira]

Hi,

I'm aware that my issue is probably not a big deal, but still, I can't find any documentation or people who seems to encounter this problem. So I'm posting here, after few hours of testing and research.

Note: ATM I'm trying this without TLS at all to simplify the process.
Note 2 : My private registry is added to the insecure registries and I disabled TLS verification, on my client-side Docker instance.

Description of my issue :
I have a private Docker Registry v2 on my local network, and I want to push an image from my local computer to this registry.

I do:
> docker tag my-image 172.19.x.x:5000/myimage-repo
And then:
> docker push 172.19.x.x:5000/myimage-repo

I got this output:
Put http://172.19.x.x:5000/v1/repositories/myimage-repo/: dial tcp 172.19.x.x:5000: getsockopt: no route to host

I tried to play a bit with url, but my Docker client-side, is trying to connect to reach something behind V1/repositories.

Versions & Specs :
I'm on Windows 10 and using Docker v17.03. The registry (v2.6.0) is a Docker container running on a VM.

How can I tell to Docker to reach registry v2 ? Thanks a lot,

[Vespira]

Tried with latest version on a Linux distribution, I could correctly push or pull images from my insecure registry, with addind the hostname and port in insecure-registries section of Docker Daemon. The issue seems to occur on Docker for Windows.

Ok also on a co-worker computer ,the push succeed. I added a cert & key to the registry Docker Image but it won't change anything on my computer.

[Vespira]

Ok, I just reinstalled Docker and it worked. I lost almost 2 days trying to understand, trying to change settings on my Docker, to modify the running registry image or it's conf... I should have tried that first, but I had to configure it again, and i'll have to build all images again but it's ok.

This state is really weird that said :(

[dnephin]

I believe that if there is a network error it will fall back to trying V1.

Are you still experiencing this problem?

[thaJeztah]

The daemon logs (in debug mode) should show which URLs are tried, and why it fell back. For docker 17.03, there's also a daemon configuration option to disable contacting legacy (V1) registries - this is the default in Docker 17.06.

Given that the issue was resolved, let's close this, but feel free to comment if you run into this again, and have more information