Enforce host networking for Ceph CSI controller plugin pods

[rohan47]

In provider mode with dedicated storage network, Where CSI is running on a client cluster both Ceph and Ceph CSI pods must run on the host network.
The Ceph CSI controller plugin pods must also run on the host network, as the dedicated network is not accessible from the pod network. Currently, the controller plugin operates on the pod network.

[obnoxxx]

similar feature for rook: rook/rook#14585

[obnoxxx]

@rohan47 , @nb-ohad , @Madhu-1 , @leelavg, Let's discuss how this should be implemented. In rook, the use of host network can now be enforced by a newly introduced boolean operator setting called ROOK_ENFORCE_HOST_NETWORK which can be set in the operator's configmap or as an environment variable. Is something similar appropriate for ceph-csi-operator?

[leelavg]

Is something similar appropriate for ceph-csi-operator

• yes, a general flow is, CSI config is surfaced by the operator based on CRDs, more info is at design link and wherever appropriate we can introduce it and have backing controller reconcile it

[rohan47]

As per the design doc, currently the network is configured via annotations but I don't see any configuration for HostNetwork @leelavg

nodePlugin:
      priorityClassName: system-node-critical
      updateStrategy:
        type: RollingUpdate
        rollingUpdate:
          maxUnavailable: 1
      labels:
        app: csi
      annotations:
        k8s.v1.cni.cncf.io/networks: macvlan-conf-1
      affinity:```

[leelavg]

This can be brought up in the meeting and propose change.

[obnoxxx]

/assign

[github-actions]

Thanks for taking this issue! Let us know if you have any questions!