Managing Docker Image Updates

[rmuraix]

Currently, this repository maintains a Conda-based Docker image designed for GPU environments. However, whenever the Dockerfile or conda_env.yaml is modified, the image must be manually rebuilt and pushed. This makes it difficult to keep the environment up to date, and as a result, the Docker image on Docker Hub may not function as expected.

Challenges

• Manual image updates are cumbersome
  • Updating the image manually for every change places a significant burden on the maintainer and increases the risk of outdated images.
• Difficulties in CI automation
  • Automating the build and push process using GitHub Actions would be ideal, but the default runners do not support GPUs, which may cause the build to fail.
  • Using external GPU-enabled runners would be a reliable solution but comes with additional costs.

Possible Solutions

To reduce the maintainer's workload, the following approaches could be considered:

• Attempt CI automation
  • There is a possibility that the image can be successfully built in a CPU environment. It may be worth testing whether automation is feasible.
• Discontinue public image distribution
  • If the pre-built image is not widely used, an alternative approach could be to stop publishing it on Docker Hub and instead encourage users to build the image themselves.

Thanks for awesome work.