RedHat Openshift certification

[rinus277]

Hi all,

I’m contacting you as a developer working for a software company that has adopted Argo Workflows as the workflow orchestrator inside our platform.
We deploy on many Kubernetes distributions, including Red Hat OpenShift, and we are facing increasing resistance from enterprise customers when installing Argo Workflows on their OpenShift clusters.

The obstacles are always the same:

Argo Workflows requires installing several cluster-scoped CRDs, webhooks and controllers.

Many OpenShift environments enforce strict multi-tenant policies and don’t allow tenants to install or manage cluster-level objects.

Since there is no certified OperatorHub entry or Red Hat–validated operator, platform teams classify Argo Workflows as “unsupported”, even though the project is CNCF-graduated and widely used.

Because of this, it is becoming very hard to adopt Argo Workflows in regulated and enterprise OpenShift environments, even when the product strongly benefits from it.

Would the project be open to working toward an enterprise-grade operator and OpenShift certification via OperatorHub?
Having an officially packaged and certified operator would remove most of the current adoption blockers and would significantly help companies like ours promote Argo Workflows inside large organizations.

If there is already any roadmap, discussion or early effort in this direction, I’d be happy to contribute testing or feedback.

Thanks for your time and for the great work on the project.

[Joibel]

We'd welcome someone willing to take on authoring and maintaining an operator.

There used to be one: https://github.com/argoproj-labs/argo-workflows-operator, but this is before my time on the project. We could create a space for something in argoproj-labs which is where all the argo operators currently live if you're interested in making something.

Using operator-sdk to develop a helm based operator would probably be pretty easy. The installation of workflows is pretty trivial and our helm charts are good.

Argo Workflows requires installing several cluster-scoped CRDs, webhooks and controllers.

All CRDs are cluster scoped by the design of kubernetes. Some of ours are non-namespaced, but optional if you're running a namespaced controller. We don't have any webhooks (and if we did they'd be optional) and there is just one, (singleton), controller (it supports leader elections if you feel like running it that way). We are not interested in developing a version of workflows that works without CRDs, but there has been some external interest in doing that but there is no POC of that even, and it would be just for testing.

If you're serious in investing time in doing this, I'm in CNCF slack (Alan Clucas) and we can chat there or have a zoom call.

[rinus277]

Hi Alan,
thank you for the clarification and for the availability to support a potential operator effort.

We have also been in contact with Red Hat regarding OpenShift certification. Their feedback is that Argo Workflows is currently distributed and supported as container images within the OpenShift AI product, but there is no certified Operator for general OpenShift usage.
They also mentioned that any initiative toward a productizable, Operator-controlled distribution would likely need to originate from the Argo project or one of its corporate sponsors, rather than from Red Hat itself.

Given this context, we are internally evaluating the feasibility and impact of contributing to an Operator in the future. It’s clear that having an enterprise-grade Operator (even initially hosted under argoproj-labs) would greatly improve adoption in regulated OpenShift environments, but we need some time to assess what this would mean for us in terms of effort and long-term maintenance.

For now, we wanted to share what Red Hat told us and let you know that we are considering the idea, even if we are not committing to it yet. We will keep you updated as soon as we have a clearer view.

Thanks again for your openness and for the support.

Best regards,
Gregorio