Virtual Machines appear as "Inaccessible" following deletion of Virtual Disk in iDRAC

Products

• VMware vSphere ESXi

Issue/Introduction

Users may experience the following symptoms on ESXi hosts (specifically Dell PowerEdge platforms):

• Virtual Machines (VMs) are listed as Inaccessible in the vSphere Client or Host Client.
• The datastore hosting the affected VMs appears in the storage list but reports a capacity and free space of 0 bytes.
• Attempts to browse the datastore fail or show no files.
• System logs or iDRAC Lifecycle Manager events indicate a specific Virtual Disk ID (e.g., ID 235) was deleted or is missing.
• The underlying storage device (NAA ID) shows a status of dead or is missing from the /dev/disks/ directory.

Environment

ESXi
Storage: Local Disks

Cause

This issue occurs when a Virtual Disk (LUN) is deleted at the hardware controller level (iDRAC/PERC) while it is still mounted as a VMFS datastore in ESXi.

When the hardware removes the target, the ESXi host loses the communication path to the storage. Because the datastore metadata exists in the host's memory but the physical blocks are unreachable, the host reports the datastore size as 0. Consequently, the configuration files (.vmx) for VMs residing on that datastore become unavailable, leading to the "Inaccessible" status.

Resolution

To resolve this issue and confirm the correlation between the deleted hardware disk and the missing datastore, follow these diagnostic steps:

1. Identify the Missing Datastore
   
   Run the following command to identify if the datastore shows zero capacity:
   esxcli storage filesystem list
   
   Example Output:
   DatastoreName... Mounted: true Type: VMFS-6 Size: 0 Free: 0
   
   
2. Verify Device Path Status
   
   Check the status of the SCSI devices. Look for devices in a dead state:
   esxcfg-mpath -b
   
   Example Output:
   naa..################################ : Local DELL Disk (naa..################################)
   vmhba3:C3:T107:L0 LUN:0 state:dead sas Adapter: Unavailable Target: Unavailable
   
   
3. Correlate iDRAC Virtual Disk ID with ESXi NAA ID
   
   Review the ESXi boot logs to map the TargetId (from the controller) to the runtime path. In the example below, TargetId 235 (from iDRAC) correlates to path C3:T107:L0:
   zcat /var/log/boot.gz | grep "mfi_update_tgt_prop"
   
   Log Snippet:
   lsi_mr3_...: mfi_update_tgt_prop: 960: Tgp Prop : C3T107 ... TargetId 235 device_id 235
   
   By matching C3:T107:L0 from the log to the dead device in esxcfg-mpath -b, you confirm that the deleted iDRAC disk (235) is the one that backed the missing datastore.
   
   
4. Recovery
   
   If the deletion was accidental: You must attempt to recover/recreate the Virtual Disk at the RAID controller level. Note that recreating a RAID volume usually results in data loss unless "Retain Metadata" options are available and supported by the controller.
   
   If the deletion was intentional:
   • Unregister the "Inaccessible" VMs from the inventory.
   • Unmount and detach the stale datastore/device from the ESXi host.
   • Perform a storage rescan: esxcli storage core adapter rescan --all