I'm part of a Kubernetes platform team that mainly uses GitOps. I'm developing a release tool and integrated it with our existing Django dashboard, utilizing Celery for some business processes. Recently, my senior colleagues suggested that we transition everything to Custom Resource Definitions (CRDs) and phase out the database. While I see the benefits of using CRDs—like having a watcher or easily accessing resources with kubectl—I'm feeling uneasy about the idea of turning everything into CRDs. Is this practice reasonable, or does it seem too extreme?
1 Answer
It's really important to understand that CRDs are not just simple data objects. They represent the state of the system and are subject to control loops depending on that state. If you're just translating Django models directly to CRDs, you're missing their actual functionality and purpose.
If we don't implement an operator, wouldn't that just make it a data object? What's the downside of using CRDs in that manner?