ESXi Host Loss of Datastore Access due to PDL or Partition Table Corruption

Products

• VMware vSphere ESXi

Issue/Introduction

Symptoms

• vCenter Server UI displays datastores as Inaccessible with 0.0B total/free capacity.
• Virtual Machines appear as "Ghost" states or Inaccessible in inventory.
• Datastore expansion or VM console operations fail with TypeError: Cannot read properties of null (reading 'scsiDiskInfos').
• vMotion migrations freeze at 0% or timeout.
• Log Errors:
  • vmkernel.log: Waiting for timed out HB
  • hostd.log: state in doubt; requested fast path state update
  • vmkernel.log: nfnic: <2>: FCPIO_DATA_CNT_MISMATCH IO failure!
  • SCSI status: H:0x8 (H_TIMEOUT)

Environment

• VMware vSphere ESXi 7.x / 8.x
• Shared Storage / SAN Environments

Cause

This issue is caused by a Storage array Permanent Device Loss (PDL) or an accidental partition table overwrite (Legacy MBR instead of GPT). When the VMFS signature is lost or the LUN is unmapped at the array level, the ESXi host cannot perform I/O, leading to the 0.0B capacity report.

Technical Reference: SCSI Sense Codes

Sense Code	Meaning	Action
0x5 0x25 0x0	Illegal Request / Logical Unit Not Supported	Check LUN mapping on Storage Array
H:0x8	Host Timeout / No Space	Investigate Storage Controller responsiveness
H:0x0 D:0x2	Check Condition	Verify physical layer (SFPs, Cabling)

Resolution

1. Analyze vmkernel.log to confirm PDL status codes.
2. Engage storage vendor to investigate LUN mapping and "PERM LOSS" status.
3. If the partition table is corrupted (MBR reported instead of GPT):
   1. Re-initialize the LUN with a clean GPT signature.
   2. Restore virtual machines from backup.
4. To safely remove or remount affected volumes, refer to:
   • Detaching a LUN from an ESXi host (KB 323128)
   • Remounting a disconnected VMFS datastore (KB 397144)

Additional Information

• For update notifications, subscribe to this article: https://knowledge.broadcom.com/external/article/275360.

• Contact Broadcom Support