Multi arch build / GHA cache

[ro0NL]

Contributing guidelines

• I've read the contributing guidelines and wholeheartedly agree

I've found a bug, and:

• The documentation does not mention anything about my problem
• There are no open or closed issues that are related to my problem

Description

We're building multi arch Dockerfile (AMD/ARM)

Using a matrix of runners we noticed the default runner (ubuntu-latest) leverages GHA cache, but our custom runner doesnt.

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

Are we missing something?

Expected behaviour

Fast build

Actual behaviour

Slow build

Repository URL

No response

Workflow run URL

No response

YAML workflow

runs-on: ${{ matrix.runner }}
        steps:
            -
                uses: actions/checkout@v4.2.2
            -
                uses: docker/setup-buildx-action@v3.8.0
            -
                id: build
                uses: docker/build-push-action@v6.13.0
                with:
                    context: .
                    load: true
                    # https://docs.docker.com/build/ci/github-actions/cache/#github-cache
                    cache-from: type=gha
                    cache-to: type=gha,mode=max

Workflow logs

No response

BuildKit logs

Additional info

No response

[crazy-max]

we noticed the default runner (ubuntu-latest) leverages GHA cache, but our custom runner doesnt.

Would need workflow logs and BuildKit logs to figure out what's going on. Just a heads-up: we don't support or test outside of GitHub-hosted runner infrastructure, though we can assist on a best-effort basis. Thanks for your understanding.

[ro0NL]

Understood :)

https://github.com/etrias-nl/php/actions/runs/13008973505
https://github.com/etrias-nl/php/blob/main/.github/workflows/docker-build.yaml

[ro0NL]

so first (due to timing) the AMD build kicks off, it's cached: #13 CACHED

then the ARM build kicked off, which is not cached.

after both builds, we start a push in a single job (platforms: linux/amd64,linux/arm64), which should leverage previous build cache.

However in the push job we noticed the latter build (ARM) is actually cached, and the AMD build isnt.

As if they are both sharing the same cache key.

[ro0NL]

will close due to unsupported features. being;

• build (multi arch, cached)
• test
• push (cached)

As if they are both sharing the same cache key.

[crazy-max]

Does it work if you put a scope matching the runner instance like:

                    cache-from: type=gha,scope=build-${{ matrix.runner }}
                    cache-to: type=gha,scope=build-${{ matrix.runner }},mode=max