Recommended way of managing Kargo resources?

[oscr]

In the Quickstart, resources such as Warehouse, Project etc are created manually with kargo apply.

However in a production scenario where we want to store the resources as code. What are your recommendations for:

1. Where should these resources be stored? In each microservice repo or in a central "kargo resources" repo?
2. What would the recommended way of reconciling the changes be? Should there be an ArgoCD instance to reconcile the changes?

[krancour]

However in a production scenario where we want to store the resources as code.

That's the right idea. The whole notion behind Kargo's moving parts being CRDs was precisely so that one could gitops their Kargo projects/pipelines.

Where should these resources be stored? In each microservice repo or in a central "kargo resources" repo?

My sense is this is really going to come down to an individual organization's requirements and preferences. Just like the question of "Do I store my application's manifests alongside its source code or in a separate repo?", I think a lot of people are going to have different opinions on this and will have various reasons for coming down on one side of it or the other.

What would the recommended way of reconciling the changes be? Should there be an ArgoCD instance to reconcile the changes?

Yes. We're clearly biased toward Argo CD. I think I would suggest something like using the "app of apps" pattern where each of the lower level Apps manages the resources of a single Kargo Project.

[oscr]

Thanks for the answer, and sorry for my somewhat slow reply 🙏

My sense is this is really going to come down to an individual organization's requirements and preferences. Just like the question of "Do I store my application's manifests alongside its source code or in a separate repo?", I think a lot of people are going to have different opinions on this and will have various reasons for coming down on one side of it or the other.

The Argo CD best practice seems to lean towards having it separate: Using a separate Git repository to hold your Kubernetes manifests, keeping the config separate from your application source code, is highly recommended for the following reasons.

If I understand this correctly, a Kargo Warehouse can have one or more git subscriptions. Which means if you have a large number of repos, then Kargo will have to continuously poll all these repos. Because it doesn't support webhooks yet(?). Is this correct? So at this point using a single repo would be a safer bet from a performance perspective.

Yes. We're clearly biased toward Argo CD. I think I would suggest something like using the "app of apps" pattern where each of the lower level Apps manages the resources of a single Kargo Project.

I think this sounds like a good approach. Maybe this is something which should added to the Kargo docs? Would you be open for a draft pr within this area?