nbviewer icon indicating copy to clipboard operation
nbviewer copied to clipboard

Support for Multiplatform Docker Image (amd64, arm64)

Open kgns opened this issue 3 years ago • 0 comments

Is your feature request related to a problem? Please describe. Currently the official image for this project is built against linux/amd64 architecture (https://hub.docker.com/r/jupyter/nbviewer/tags). With the rise of ARM CPU usage on the cloud, open source projects are starting to utilize multiplatform docker builds to generate multiple images per supported CPU architecture. We are using AWS EC2 instances based on ARM CPUs for our deployments. This means we cannot use nbviewer official docker image as it is, since it only works on amd64 arch.

Describe the solution you'd like Docker has an experimental project called buildx that allows building multiplatform images for different architectures utilizing QEMU virtualization. While experimental, it works flawlessly once set up correctly. I searched this repo to see where the docker build pipeline was, but could not find it. My solution suggestion would be to update the pipeline to use docker buildx instead of docker build

Describe alternatives you've considered Not so much as an alternative, but more of a workaround is for everyone to build their own multiplatform image for this project and push it to an either public or a private container repository and use that image on production, but this would mean extra maintenance for every user whereas it could have been handled at a single point

kgns avatar Apr 14 '21 10:04 kgns