Investigate use of ArgoCD pull model in place of ACM Application/Subscription

[grdryn]

It sounds like ArgoCD is the future for our use-case in ACM, so this issue is to investigate that and try to figure out what changes would be required for us to make that change now.

[adelton]

@grdryn What is the functional goal of this investigation? Just use a different technology? Or make sure that the near-edge setup works even if the ACM hub cluster cannot initiate connection to the managed clusters? Or something else?

[grdryn]

@adelton when we start to integrate this work with RHODS, we should try to have it using the recommended approach from the beginning if possible. For edge scenarios, especially more constrained k8s environments using MicroShift, this ArgoCD pull model will apparently be the recommended approach in ACM rather than the Application/Subscription approach that we used in the PoC. This investigation is just me trying to get an understanding of it and how it differs, and how much work might be involved in switching to that approach, if that's what we want to do. Does that make sense?

Here's a related Jira issue: GITOPS-3374

[adelton]

@adelton when we start to integrate this work with RHODS, we should try to have it using the recommended approach from the beginning if possible.

For near-edge cases, would you have a link to where this recommendation comes from?

For edge scenarios, especially more constrained k8s environments using MicroShift, this ArgoCD pull model will apparently be the recommended approach in ACM rather than the Application/Subscription approach that we used in the PoC. This investigation is just me trying to get an understanding of it and how it differs, and how much work might be involved in switching to that approach, if that's what we want to do. Does that make sense?

So this investigation is based on the anticipated stretching out to the realm of far-edge, not that the current ACM approach would be wrong or not recommended per se. Do I get it right?

I try to grasp the context of this issues / work since I've been looking into the far edge situation (even beyond MicroShift) and currently ArgoCD will get us only so far.

Here's a related Jira issue: GITOPS-3374

Thanks, this is useful.

[LaVLaS]

I try to grasp the context of this issues / work since I've been looking into the far edge situation (even beyond MicroShift) and currently ArgoCD will get us only so far.

The use of ACM (w/ ArgoCD) for *Edge scenarios will only apply to OpenShift edge environments...assuming there is no resource or environmental constraint that blocks an ACM Agent w/ ArgoCD from managing the MicroShift environment...which would be a bug or enhancement we could contribute back to ACM.

[grdryn]

Closing this as follow-up issues have been created in Jira:

• RHOAIEDGE-44
• RHOAIEDGE-45