docker login error

Accelerated Computing NGC GPU Cloud Docker and NVIDIA Docker
#1

I’m trying to log in to nvcr.io and keep getting an error.

What I’ve done:

Then I try to login using:

docker login nvcr.io
Username: $oauthtoken
Password:

The response I get is:

Error response from daemon: Get https://nvcr.io/v2/: proxyconnect tcp: tls: oversized record received with length 20527

If I do not properly set up http-proxy.conf I get a timeout error instead. This makes me think that a connection is being made but still somehow failing.

Any suggestions are appreciated.

Thanks,
Eric

#2

Eric,

I suspect you are correct that you are dealing with a network / proxy issue. Unfortunately, troubleshooting that remotely is more than what we can do on this forum.

#3

The proxy works for literally everything else on my machine. There’s seriously nothing anybody can recommend I try in terms of configuring nvidia-docker correctly to use it?

#4

In case anybody else runs into this issue, here’s what I did to solve it (following advice from https://stackoverflow.com/questions/50717390/ubuntu-docker-proxyconnect-tcp-tls-oversized-record-received-with-length-20)

I created /etc/systemd/system/docker.service.d/proxy.conf with the following three lines:

[Service]
Environment="HTTP_PROXY=http://wwwproxy.mysite.com:80"
Environment="HTTPS_PROXY=http://wwwproxy.mysite.com:80"

Notice that the HTTPS_PROXY points to an address of the form http:// rather than https://

I can’t say I understand why this is the case. Everything else on my system that needs and HTTPS_PROXY set uses an https:// address for the proxy server.

After you’ve created this file, perform the following to reload the daemon configuration and restart docker:

sudo systemctl daemon-reload
sudo systemctl restart docker.service