Avoid generic name for Cluster wide resources

[Jean-Daniel]

The deployment manifest (v0.5) creates cluster wide resources (ClusterRole, ClusterRoleBinding, …) with generic name like metrics-auth and metrics-reader.

Using generic name for cluster wide resource is discouraged as they may conflict with other components deployed in the same cluster.

These resources name should probably be prefixed by barman-plugin- to avoid conflict with other cluster components.

Even the ClusterRole like objectstore-viewer-role should make it clear they are barman-objectstore and not related to any other objectstore (like CephObjectStore).

And even for the namespaced resources (leader-election-rolebinding, selfsigned-issuer) it may be an issue. If you try to deploy 2 CNPG plugins that both require specific leader-election-rolebinding, they may clash.

[mnencia]

Thank you for raising this.

You're absolutely right — using generic names for cluster-wide resources increases the risk of naming conflicts, especially in shared environments or when deploying multiple components with overlapping responsibilities.

It would be a good improvement to introduce a consistent prefix such as plugin-barman- for all cluster-wide resources (ClusterRole, ClusterRoleBinding, etc.), and possibly apply similar namespacing to namespaced resources (e.g., leader-election-rolebinding, selfsigned-issuer) to prevent potential clashes in multi-plugin scenarios.

As this would be a breaking change, any implementation should carefully assess the impact on existing deployments and ensure clear upgrade guidance. We’d be happy to review a contribution in this direction.