Storage

User volumes are PersistentVolumeClaims created by the operator of whichever cluster a session lands on, from the StorageClass given to that operator as operator.volumeStorageClass. gShare itself never touches the storage; it registers pools, admits volumes against their capacity, and pins each volume to the cluster where it was first mounted.

The reference setup

The reference installation is one ZFS + NFS storage node served to every cluster through democratic-csi:

• cluster-bootstrap.sh installs the driver on the control-plane cluster with the values in deploy/storage/democratic-csi-values.yaml, producing the gshare-data StorageClass;
• every node needs an NFS client (nfs-common / nfs-utils) and network reach to the storage box on 2049/tcp;
• the operator is given --volume-storage-class gshare-data.

The step-by-step node preparation is in Cluster setup → Storage node. Any other CSI driver works the same way as long as it provisions ReadWriteMany volumes that every node of the cluster can mount.

Registering a pool

A pool is a registered object, not something inferred from node roles. Register it in the console under Resources → Volumes → Storage pools (Administrator guide → Storage) or with POST /api/v1/storage/pools:

Field	Meaning
cluster_id	The cluster whose storage server this is.
storage_class	The StorageClass that provisions from it — the same name the operator was given.
share_scope	all — every cluster in the fleet may place volumes on it (one NFS box exported to all, the usual shape); selected — plus the list of allowed clusters.
manual_capacity_gb	The pool's size, used when the driver does not publish capacity.

Several storage servers are several pools. A volume lives on exactly one — the StorageClass its PVC names decides where — so the capacity gate and the dashboard take the largest usable pool as the bound, never the sum.

Where the capacity number comes from

1. csi — the driver's GetCapacity, published by the external-provisioner as CSIStorageCapacity objects and read by the operator on its volume-sync tick. The only automatic source. The provisioner needs --enable-capacity and --capacity-for-immediate-binding=true, and the driver must report node topology; without topology the objects churn and nothing usable is published.
2. manual — the figure stated on the pool. STORAGE_POOL_CAPACITY_GB (storage.poolCapacityGb in the chart) still works as a fleet-wide default.
3. node_disk — the storage node's system drive, labeled as a stand-in.

The dashboard's storage tile names the source it used.

Attached clusters

An attached cluster needs the same driver pointed at the same storage box. Export the driver's values from the control plane and pass them to attach-cluster.sh:

helm -n gshare-storage get values gshare-storage > csi-values.yaml && chmod 600 csi-values.yaml
# edit controller.nodeSelector → a node of the new cluster
./hack/attach-cluster.sh ... --storage-values csi-values.yaml

The values file carries the storage box's SSH key: keep it out of the repository and delete it when the attach is done. See Multi-cluster.

Reclamation

A deleted volume's PVC is kept for api.volumeReclaimGraceHours (default 24) before the operator deletes it and the driver destroys the dataset. Set it to 0 to reclaim on the next sync.