Kubernetes troubleshooting · Lesson 7 of 11 · 3 min read
Storage problems
PVC stuck Pending
kubectl describe pvc data-orders-0 -n shop
kubectl get storageclass
- No default StorageClass, or the named class doesn't exist.
- Provisioner (EBS CSI driver) not installed, or its controller lacks IAM permissions (IRSA/Pod Identity) → events show
failed to provision volume. WaitForFirstConsumer— PVC stays Pending until a pod uses it (that's normal).- Requested access mode unsupported (EBS can't do
ReadWriteMany— use EFS).
Enlarge diagram
Swipe sideways to see all of it, or tap to open it full screen.
Multi-Attach and AZ issues
Multi-Attach error for volume ... Volume is already exclusively attached to one node— an RWO EBS volume is still attached to the old node (node died, pod rescheduled). Usually resolves after the attach/detach controller times out (~6 min); delete the old pod/VolumeAttachment if stuck, and avoid RWO volumes in Deployments with >1 replica.- Pod with an EBS PVC can't move to another AZ — EBS is zonal. After an AZ failure, the pod can only run in the volume's AZ. Use topology-aware scheduling, keep capacity per AZ, and for HA rely on application-level replication (databases) rather than moving disks.
Enlarge diagram
Swipe sideways to see all of it, or tap to open it full screen.
Volume permissions and capacity
permission deniedwriting to a mounted volume →securityContext.fsGroupso the kubelet sets group ownership, or an init container tochown.- Volume full →
kubectl exec ... df -h; expand the PVC (allowVolumeExpansion: true, editspec.resources.requests.storage).