Storage Nodes
A storage node is described by its StorageNode resource, one per worker and NUMA socket slot, and operated through
StorageNodeOps resources. Each StorageNodeOps names the target StorageNode, carries one action, and is driven to
completion by the Simplyblock Operator. Two operations are raised by the operator itself: a HostMaintenance when the
worker of a node is cordoned, and a Remove when a StorageNode that holds data is deleted.
A single device of a node is restarted with a StorageDeviceOps resource, as described in
Storage Node Actions.
Topics
| Topic | Purpose |
|---|---|
| Storage Node Actions | The StorageNodeOps kind: the target node, the actions, the locking, the tracking. |
| Shutting Down a Storage Node | Stopping one node while the rest of the cluster keeps serving I/O. |
| Restarting a Storage Node | Stopping and starting one node, optionally forced and reattaching its volumes. |
| Suspending a Storage Node | Stopping new volumes from being placed on a node that keeps running. |
| Resuming a Storage Node | Returning a suspended node to normal service. |
| Replacing a Storage Node | Choosing between a migration and a removal when hardware is replaced. |
| Migrating a Storage Node | Moving a storage node onto a different worker, keeping its identity and devices. |
| Removing a Storage Node | Draining the volumes off a storage node and taking it out of the cluster. |
| Coordinated Worker Node Drain | The HostMaintenance operation that protects storage while a worker is cordoned and drained. |