DGX Spark Users unable to pull from nvcr.io

Accelerated Computing DGX Spark / GB10 User Forum DGX Spark / GB10
1

I am Spark user in China.

Docker unable to pull containers from nvcr.io.

I have API token.

docker run --rm --gpus all nvcr.io/nvidia/tensorrt-llm/release:spark-single-gpu-dev nvidia-smi

Seem like it is blocked in China.

Are there any mirror sites I can use?

Thanks!

Mark

2

What is the error you get? You may need to add your user to the docker group first.
If you see an error when running docker ps then

sudo usermod -aG docker $USER
newgrp docker

Make sure to login first with

export NGC_API_KEY="<YOUR_NGC_API_KEY>"
echo "$NGC_API_KEY" | docker login nvcr.io --username '$oauthtoken' --password-stdin
3

I am a DGX Spark user in the United States and am also experiencing this issue. Docker is configured to run appropriately and I have been running other containers without issue.

docker login ncvr.io
Username: $oauthtoken
Password:
Error response from daemon: Get “https://ncvr.io/v2/”: dial tcp: lookup ncvr.io on 127.0.0.53:53: no such host

At first I thought that this was a name server issue, but IP address lookup for ncvr.io appears to work as expected.

nslookup nvcr.io
Server:		127.0.0.53
Address:	127.0.0.53#53

Name:	nvcr.io
Address: 54.183.242.120

When I ping the IP address, it appears to hang, though.

ping nvcr.io
PING nvcr.io (54.183.242.120) 56(84) bytes of data.
^C
— nvcr.io ping statistics —35 packets transmitted, 0 received, 100% packet loss, time 34819ms

Any suggestions to resolve this problem would be appreciated.

4

Hi, I tried

export NGC_API_KEY=“<YOUR_NGC_API_KEY>”
echo “$NGC_API_KEY” | docker login nvcr.io --username ‘$oauthtoken’ --password-stdin

what I received is 403 error.

However when I use my singapore SIM card as roaming internet I can log into nvcr.io. But due to the size of the images I cannot afford to use my roaming data to download the images.