The following symptoms may occur after unmounting or deleting a datastore if the underlying LUN was unmapped from the storage array without a prior "Detach" operation in vSphere:
alarm.StorageConnectivityAlarm for a decommissioned device.vmkernel.log reports: WARNING: ScsiDevice: ####: Device :naa.### has been removed or is permanently inaccessible.hostd.log shows the host attempting to initialize defunct devices cached in the ConfigStore.VMware vSphere ESXi 8.x
VMware vSphere ESXi 9.x
This issue occurs when the Detach step is omitted from the storage decommissioning workflow. While unmounting removes the filesystem, the ESXi kernel continues to poll the SCSI block device. If the LUN is unpresented array-side, ESXi registers a Permanent Device Loss (PDL) and caches the stale device identifier in the ConfigStore database.
If possible reboot the ESXi hosts in the cluster to clear stale devices from host configurations. If host can not be rebooted Contact Support to resolve this issue without rebooting the ESXi host, the stale entries must be manually purged by Broadcom Support from the ConfigStore.
To prevent this in the future ensure to follow the below steps when decommissioning a datastore:
Evacuate VMs / Data.
Unmount Datastore in vSphere Client.
Detach Storage Device (Host > Configure > Storage Devices > Detach).
Delete / Remove Datastore from Inventory.
Unpresent LUN on Storage Array.
Perform Rescan Storage on clusters/hosts.
Contact Support If the alarm persists after performing these steps, Contact Support.