[release-1.1]/[release-1.2] Prow GCR build is slow

[wongma7]

/kind bug

What happened?

https://prow.k8s.io/job-history/gs/kubernetes-jenkins/logs/aws-ebs-csi-driver-push-images timeouts

For comparison, the github action builds take about 15 minutes.

Possible mitigations, they r not mutually exclusive

1. Stop building the debian variant, only build amazonlinux variant. This was planned for 1.3.x but we can "backport" this change to 1.1.x and 1.2.x branches too. It should halve the build time.
2. Replace GCR prow build with public ECR GitHub actions build. This way we will only need to maintain one build pipeline, github actions. I think we can't simply stop pushing to GCR* without first offering an alternative registry like public ECR that allows unauthenticated pulls (because it is useful for testing or for those who don't want to register/authenticate just to pull an image ) and that doesn't throttle (unlike Docker Hub), so if we want to stop pushing to GCR via Prow we will need to push to public ECR via GitHub actions . cc Replace GCR build with public ECR #1011

What you expected to happen?

How to reproduce it (as minimally and precisely as possible)?

Anything else we need to know?:

Environment

• Kubernetes version (use kubectl version):
• Driver version:

[wongma7]

https://prow.k8s.io/view/gs/kubernetes-jenkins/logs/aws-ebs-csi-driver-push-images/1435723992483762176

#21 [linux/arm64 builder 4/4] RUN make
#21 DONE 3424.7s

takes over an hour on prow but ~10m on github.

https://github.com/kubernetes-sigs/aws-ebs-csi-driver/actions/runs/1215072462

#14 [linux/arm64 builder 4/4] RUN make
#14 DONE 564.3s

Either way it is slow I think because we are building for arm using arm virtualization when we could be building for arm without virtualization (GOARCH)

[wongma7]

Test started today at 10:40 AM passed after 1h4m6s. (more info)

https://prow.k8s.io/view/gs/kubernetes-jenkins/logs/aws-ebs-csi-driver-push-images/1436021635202158592