I follow the tutorial on this git GitHub - SvenBrnn/runpod-worker-ollama: A serverless ollama worker for runpod.io and I get around 10 Gib image because I need to include a LLM model to the images. It should be easy to upload with my internet connection. But the thing is, the upload speed is kind of random. One time I spent 3 hours and barely pushing 300Mb, another time I’m able to complete it in 2 hours. I have tested my internet speed and it’s around 20Mbps for upload. It should be enough but somehow it stuck when pushing image.
I have read some issue on slow speed when push to hub, and I read about using Ipv4 only, but it doesn’t seems to work
So is there a trick to push large image to hub faster?
If Docker Hub’s servers are busy or there is a high network traffic, I guess it could change the speed of the upload, especially for large images. There is probably no trick, if you need a large image, but you can try to minimize the number of layers. I’m not sure that helps though and not just makes it worse. I never had to push such a large image to Docker Hub, only to private registries.
Do you also have a subscription or using a free account? I don’t know if that matters, but maybe.
that seems to be extremely slow as 300MB images are not rare, so it is also possible you saw that speed when there were issues on Docker Hub.
If you experience that again or the push command fails too, you can report it on GitHub