VMware: LU reset by VMware ESXi

Storage: VNX
Software: PowerPath/VE

When a reservation conflict takes place, ESXi issues a LU reset.
A LU reset releases reservations established by the reserve or release method and removes all tasks for all initiators for the addressed logical unit.
Also a LU reset (and a Bus reset) clears ***all*** Group Registrations and Reservations for PowerPath in VNX.

You may see the 05/80/20 alert message on the other ESXi host that did not issue a LU reset as follows.

2016-06-12T21:16:26.183Z cpu34:33711)ALERT: PowerPath:MpxRegister: PRO service action 46 failed, 0x5/0x80/0x20 for vmhba1:C0:T3:L5

UseDeviceReset    UseLUNReset    Type of Reset
      0                0            Bus Reset
      0                1            LU Reset
      1                0            Device Reset
      1                1            LU Reset (default)

Storage: Symmetrix

A LU reset is often observed by rebooting a VM if it is Windows with an RDM (see https://kb.vmware.com/kb/2033174).
You may see the message below on the other ESXi host that did not reboot a VM.

2016-02-10T16:05:52.769Z cpu12:32980)WARNING: NMP: nmpDeviceTaskMgmt:2266: Attempt to issue lun reset on device naa.60000970000292604067533033393637. This will clear any SCSI-2 reservations on the device.

When a LUN reset occurs, a 013D is logged then it cancels all I/Os on the device on that I-T-L nexus (that is, hba -> port -> lun). It also releases the device lock if its currently locked, and sets the UA_RESET Unit Attention on ***all other I-T-L***. The next I/O the device on each I-T-L fails with a check condition and sense 06/29/00. This is an informational error, and the host can or should retry automatically. You can disable it by setting the Avoid Reset On Broadcast flag.