Apply caching best practices to all images

[ianpittwood]

We currently do not share or invalidate caches efficiently in most of our images which causes builds to take longer and for cache bloating to take up a lot of disk space. We should follow best practices as best we can for every image to help ensure builds are efficient and that data can be shared where possible across different versions of an image.

• Caching Optimization - https://docs.docker.com/build/cache/optimize/
• Cache Mounts - https://docs.docker.com/reference/dockerfile/#run---mounttypecache
• Cache Invalidation - https://docs.docker.com/build/cache/invalidation/
• GHA Cache Backend - https://docs.docker.com/build/cache/backends/gha/

Pull Requests

• Use GHCR repo as image variant build cache in GHA #245

[bschwedler]

We have a few different options for the layer that we choose for layer caching

Python on Whales

Use the cache_from and cache_to parameters when calling bake() or build().

This could be the cleanest way, but we would need to be able to determine effectively whether we are running in GHA in order to use the gha cache type.

Render in bakefile

We could have bakery render the cache-from and cache-to in the ephemeral bakefile.

This would apply both locally and in CI, but would not necessarily translate into other image builders.

GitHub Actions

Use the docker/bake-action and specify cache-from and cache-to

This can be specified down to the image level:
https://github.com/rstudio/partnership-snowflake/blob/239c2e99ed2ab8958d56cbba9fcdb3ed427fffdf/.github/workflows/test-posit-team.yaml#L46-L55

But this somewhat moot if we do not invoke this action directly.

[bschwedler]

We plan to take the approach of setting the parameters in the Python on Whales calls and being able to set with a cli flag or environment variable.

[bschwedler]

We are taking some motivation from https://github.com/int128/docker-build-cache-config-action but are currently only specifying one source and one destination.

[bschwedler]

After implementing cache to and from on a per-target basis:

It appears we have more work we can do on improving the caching of the connect image.

Product	Before	After
workbench
connect

[bschwedler]

Caching is now happening by default when using the shared workflow:

• Use GHCR repo as image variant build cache in GHA #245

Moving this out of In Progress so that we can address the image definitions themselves to improve the caching.