Issue: performance of Docker builds in workflow `Publish dev (Docker image of master)`

[erikmd]

Related user(s):

@erikmd

Related issue(s) or PR(s):

• Task 611 docker images #614
• Task: port multi-platform support from deploy-master.yml to release.yml #626

Related project scope(s):

build

Bug description:

As a side-effect of building multi-platform images using github's ubuntu/amd64 runners, we get a 9.1x slowdown factor, namely:

• https://github.com/ocaml-sf/learn-ocaml/actions/runs/17341306050 (11'25" for that run, 3 days ago)
• https://github.com/ocaml-sf/learn-ocaml/actions/runs/17407215422 (1h45' for today's run)

To reproduce:

• Reproducible by pushing new commits in master

Expected behavior:

No response

learn-ocaml --version

No response

git describe --long --always --abbrev=40 --tags

No response

What OS do you use?

GNU/Linux

What OS version/distribution do you use?

Ubuntu

What browser(s) do you use with learn-ocaml?

No response

What browser(s) version did you used to reproduce the issue?

No response

Screenshots (if need be):

No response

Additional context:

• Fortunately, the new time for this GHA job is not prohibitive (1h45' < maximum threshold for GHA jobs = 6h)
  and it only deals with deployment jobs (for master and tags)
• Yet, some improvements to reduce this slowdown might be considered, see e.g. the following source:
  https://depot.dev/blog/multi-platform-docker-images-in-github-actions

[erikmd]

Food for thought - as suggested in https://depot.dev/blog/multi-platform-docker-images-in-github-actions ,

docker/build-push-action supports these fields:

• cache-from: "type=gha"
• cache-to: "type=gha,mode=max"

and according to https://docs.docker.com/build/cache/backends/#cache-mode ,

While min cache is typically smaller (which speeds up import/export times, and reduces storage costs), max cache is more likely to get more cache hits. Depending on the complexity and location of your build, you should experiment with both parameters to find the results that work best for you.

[erikmd]

Comparing that run (yesterday, 1h45 of elapsed time) and this run (today, 4min of elapsed time), it appears the overhead is well mitigated using Docker caching.

So let's close this issue now.