🌱 Cross-compile container image binaries from host platform

[astefanutti]

Summary

The compilation of the binaries in the builder stage of the container image is very slow when QEMU is used:

#44 [linux/amd64 builder 11/11] RUN --mount=type=cache,target=/root/.cache/go-build     --mount=type=cache,target=/go/pkg/mod     make
#44 881.5 ln -sf kubectl-workspace bin/kubectl-workspaces
#44 881.5 ln -sf kubectl-workspace bin/kubectl-ws
#44 DONE 882.2s

#49 [linux/arm64 builder 11/11] RUN --mount=type=cache,target=/root/.cache/go-build     --mount=type=cache,target=/go/pkg/mod     make
#49 5526.2 ln -sf kubectl-workspace bin/kubectl-workspaces
#49 5526.3 ln -sf kubectl-workspace bin/kubectl-ws
#49 DONE 5527.0s

#50 [linux/ppc64le builder 11/11] RUN --mount=type=cache,target=/root/.cache/go-build     --mount=type=cache,target=/go/pkg/mod     make
#50 5266.1 ln -sf kubectl-workspace bin/kubectl-workspaces
#50 5266.4 ln -sf kubectl-workspace bin/kubectl-ws
#50 DONE 5266.9s

This PR switches to using Golang cross-compilation instead of QEMU, by forcing the builder stage to run the same platform as that of the host one, that gives:

#27 [linux/amd64 builder 11/11] RUN --mount=type=cache,target=/root/.cache/go-build     --mount=type=cache,target=/go/pkg/mod     make OS=linux ARCH=amd64
#27 963.1 ln -sf kubectl-workspace bin/kubectl-workspaces
#27 963.2 ln -sf kubectl-workspace bin/kubectl-ws
#27 DONE 963.8s

#28 [linux/amd64->arm64 builder 11/11] RUN --mount=type=cache,target=/root/.cache/go-build     --mount=type=cache,target=/go/pkg/mod     make OS=linux ARCH=arm64
#28 987.2 ln -sf kubectl-workspace bin/kubectl-workspaces
#28 987.3 ln -sf kubectl-workspace bin/kubectl-ws
#28 DONE 989.2s

#26 [linux/amd64->ppc64le builder 11/11] RUN --mount=type=cache,target=/root/.cache/go-build     --mount=type=cache,target=/go/pkg/mod     make OS=linux ARCH=ppc64le
#26 991.1 ln -sf kubectl-workspace bin/kubectl-workspaces
#26 991.1 ln -sf kubectl-workspace bin/kubectl-ws
#26 DONE 991.1s

It also triggers the image build workflow on pull requests, so non-regression of the container image build is covered, though the result image is not pushed to any registry.

[sttts]

/retest

[sttts]

is this just for testing?

[astefanutti]

I'd be inclined to keep it, so as to guarantee PRs do not break the image build for some reasons. In that case, the image is only built, not pushed, which I think is what we want for PRs:

kcp/.github/workflows/kcp-image.yaml

Line 54 in c14bdee

push: ${{ github.event_name != 'pull_request' }}

[sttts]

With 21min it is much faster now. So I am less concerned this blocks our github action queue.

[sttts]

Nice improvement.

[sttts]

/retest
/lgtm
/approve

[openshift-ci]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: sttts

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [sttts]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[ncdc]

Flake #2501