NFS server

[seanho00]

Steps

• Use CephFilesystem CRD to define backing pools for data + metadata
• Use CephNFS CRD to configure Ganesha servers
  • Specify pool to store Ganesha's own metadata; e.g., reuse metadata pool from CephFS
• Tell Ceph dashboard which pool+ns to find Ganesha config
• From Ceph dashboard, define export

Exposing NFS service

Ganesha deploys multiple services (kubectl get -n rook-ceph svc -l app=rook-ceph-nfs), each with ClusterIP, for active-active. Use host networking and haproxy outside cluster?

[seanho00]

• Wait for backport fix for name mismatch: ceph nfs cluster ls
• Define Exports
• Migrate data
• Reduce/remove drives from btrfs array
• RBD tuning
  • rbd_read_from_replica_policy: localize
  • rbd_default_order: 12
• Client config, sysctl tweaks