Why is Redpanda Operator bundling FluxCD?

[j0057]

I'm evaluating Deployment > Kubernetes > Overview > Helm and RedPanda Operator and I'm concerned about how this is supposed to work. Bundling FluxCD with Redpanda Operator seems like a big and unnecessary coupling of RedPanda to a specific GitOps framework.

It causes either duplication of resources in a cluster where FluxCD is already installed, or to cause version conflicts between the bundled and the already-installed FluxCD, and to make upgrades of either FluxCD or RedPanda Operator more complicated, and can only create trouble for users of other GitOps frameworks.

The operator should just be some CRDs and controllers for reacting to them, and have trust that deployment lifecycle of the operator itself is managed on some other layer, be it manual application of Helm charts, a GitOps framework, or Operator Lifecycle Manager, Operator-Controller, or whatever else the cluster admin uses to manage their operators.

Please consider providing a more bare-bones deployment of the operator, with no bundled GitOps tooling.

[david-yu]

Hi @j0057 thank you for the feedback. We are moving towards an operator that will remove FluxCD entirely and are targeting a release before the end of this year that coincides with 24.3.x of Redpanda Core. This has been on our radar for some time (see redpanda-data/redpanda-operator#91) as the intended benefits of using FluxCD within an operator for de-duplicating work across Helm and Operator has not been realized. We agree that having an Operator without any bundled GitOps tooling is the correct approach.