Why does Kratix write to state store?

[joebowbeer]

Please excuse the naive question, but I read that Kratix leverages GitOps (which syncs state of resources from git) and I also read that Kratix writes to the Git (or S3) state store.

Can Kratix work in a purely read-only fashion, e.g., obtaining desired state from Git and reconciling it on k8s clusters?

Relates #346

[kirederik]

Hey @joebowbeer

Kratix can orchestrate workload across multiple destinations, and it does that via GitOps--and that's why it needs to write to the state store.

The diagram below explains the "standard" installation of a Platform with Kratix:

As you can see, users send request to the Platform cluster, that is picked up by the Promise which runs its Workflows and output some documents. Those documents are written to the state store by Kratix. On the other side, a GitOps agent (like Flux) is running, and reconciling on the documents in the state store.

You could write the requests to a state store, and have Flux/ArgoCD reconcile that on the Platform cluster -- instead of giving users direct kubectl access. Kratix does not watch the state store.

I think what you are asking is whether Kratix can schedule documents to the cluster where Kratix itself is running? If that's the case, not really. Currently, even to communicate with its own cluster, it will need to write to the state store and rely on an agent to sync the state.

I'm curious to hear what's your use case and what you have in mind!