In Kubernetes or cloud-native environments multi-attach issues occur when a volume that supports single-node access is attempted to be mounted by multiple nodes/pods simultaneously.
#As you can see below error will get during sync:
Failed to mount images in ImageGroup 'ImgGrp1_OCI_PSH_Cluster_default_data [ ERROR: Failed to mount image swift-image-data-elasticsearch-data-1. ERROR:Failed to mount /pool-01/mnt/swift-zfs-volumes/JobId-1573016-swift-volume-ImgGrp1_OCI_PSH_Cluster_default_data.swift-image-data-elasticsearch-data-1/data-elasticsearch-data-1.img. Error: Invalid argument ]
# To successfully complete the sync in SWIFT, simply delete the image group and target namespace, then re-run the same sync. The issue should not occur again afterward.
#How to identify Mutli-attach issue:
1. check the pods event:
kubectl describe pod <pod-name> -n <namespace>
check the warning as below:
Warning FailedAttachVolume Multi-Attach error for volume "pvc-xyz": volume is already attached
2. Check the PVC status, if they already Bound state or not?
kubectl get pv | grep Bound
#How to avoid Multi-attach issue:
1. Use RWX (readwritemany) volume if needed. Avoid using ReadWriteOnce (RWO) volumes in scenarios where multiple pods across nodes will need the volume.
2. Make sure the source pod using the volume is completely terminated before synching.
3. In DR scenarios, always detach volumes from the source cluster before attaching to the target.