[backend] <Image gcr.io/ml-pipeline/workflow-controller:v3.3.8-license-compliance is not supported for linux/arm64 paltform>

[Monokaix]

Environment

• How did you deploy Kubeflow Pipelines (KFP)?

while ! kustomize build example | kubectl apply -f -; do echo "Retrying to apply resources"; sleep 10; done

• KFP version:

v1.8.0

• KFP SDK version:

Steps to reproduce

1、execute while ! kustomize build example | kubectl apply -f -; do echo "Retrying to apply resources"; sleep 10; done
2、 failed to run image gcr.io/ml-pipeline/workflow-controller:v3.3.8-license-compliance , it show exec format error

Expected result

gcr.io/ml-pipeline/workflow-controller:v3.3.8-license-compliance run successfully

Materials and Reference

---

Impacted by this bug? Give it a 👍.

[Monokaix]

Is there any plan to build this image on linux/arm platform?

[chensun]

workflow-controller is based on Argo's image and we only add licenses info to it:

pipelines/third_party/argo/Dockerfile.workflow-controller

Line 16 in 609c637

FROM docker.io/argoproj/workflow-controller:${TAG}

You'll need to raise the ARM support request to https://github.com/argoproj/argo-workflows

[AndersBennedsgaard]

@chensun Argo Workflows build their images for arm (see https://github.com/argoproj/argo-workflows/blob/669d4a4e986b359dde3cd07e5616be8f713fe23d/.github/workflows/release.yaml), but the custom image created for KFP (https://github.com/kubeflow/pipelines/blob/609c637811e46eab2d5c9a915913a1520c979a88/third_party/argo/Dockerfile.workflow-controller) is built on default x86-based GH runners, so the custom image does not support arm64. If we want to create an image which supports both x86 and arm64, then we would need to change the pipeline to support this.

@Monokaix you could just use the official Argo Workflows container instead of the one built for KFP, but you will probably hit other problems when you have gotten that up and running. I for example run a cluster with a full installation of KFP v2 using the upstream Argo Workflows image, but it also takes building all KFP images outselves, as mentioned in #10309.

We might want to close this issue in favor of #10309 and add gcr.io/ml-pipeline/workflow-controller to the list of images that needs arm64-support

[github-actions]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[github-actions]

This issue has been automatically closed because it has not had recent activity. Please comment "/reopen" to reopen it.