A network-attached volume exposes storage over the network, allowing a stateful workload’s volume to move from one node to another after node failure. This differs from a local volume, which remains tied to the failed node. section 16.2.2
During StatefulSet recovery, volume location affects scheduling and startup:
- A replacement pod can be scheduled on another node when the volume is network-attached. section 16.2.2
- The pod may remain in ContainerCreating if the volume cannot detach from the failed node. section 16.2.2
- A local volume instead creates a node-affinity conflict when the original node is unreachable. section 16.2.2
Diagnosing a failed reattachment
After force-deleting quiz-1 on a failed node, the replacement pod may reach ContainerCreating on another node. kubectl describe reports: “Multi-Attach error ... Volume is already exclusively attached to one node and can't be attached to another.”
