Working with Permanent Device Loss

With earlier ESX/ESXi releases, an APD (All Paths Down) event results when the LUN becomes unavailable. The event is difficult for administrators because they do not have enough information about the state of the LUN to know which corrective action is appropriate.

In ESXi 5.0, the ESXi host can determine whether the cause of an All Paths Down (APD) event is temporary, or whether the cause is permanent device loss. A PDL status occurs when the storage array returns SCSI sense codes indicating that the LUN is no longer available or that a severe, unrecoverable hardware problem exist with it. ESXi has an improved infrastructure that can speed up operations of upper-layer applications in a device loss scenario.

Important Do not plan for APD/PDL events, for example, when you want to upgrade your hardware. Instead, perform an orderly removal of LUNs from your ESXi server, which is described in Detaching Devices and Removing a LUN, perform the operation, and add the LUN back.

To Remove a PDL LUN

How you remove a PDL LUN depends on whether it was in use.

■	If the LUN that goes into PDL is not in use by any user process or by the VMkernel, the LUN disappears by itself after a PDL.

■	If the LUN was in use when it entered PDL, delete the LUN manually by following the process described in Detaching Devices and Removing a LUN.

To Reattach a PDL LUN

1	Return the LUN to working order.

2	Remove any users of the device.

You cannot bring a device back without removing active users. The ESXi host cannot know whether the device that was added back has changed. ESXi must be able to treat the device similarly to a new device being discovered.

3	Perform a rescan to get the device back in working order.