Issue:

After reboot, HDS USP LUNs were in error state:

c2t4d4       auto            -            -            error
c2t4d5       auto            -            -            error
c2t4d6       auto            -            -            error
c2t4d7       auto            -            -            error
c2t5d0       auto            -            -            error
c2t5d1       auto            -            -            error
c2t5d2       auto            -            -            error
c2t5d3       auto            -            -            error

As a workaround,  "vxdctl enable" would change the state to "online".

After the system came up with vxconfigd dadmon disabled, SCSI inquiry through all devices on the system was working (confirmed with diskinfo and vxdmpinq). Then, vxconfigd was started (debug level 9), the problem reproduced.  It resulted in lots of disabled dmpnode notifications on the console. Disabled dmpnodes were the disks which were shown in error state in 'vxdisk list' output. There were few devices which were online also.

Interestingly, all of the online devices were not having any SCSI-3 PR keys on them. All the devices in error state were in use as part of
some fencing enabled disk groups on other nodes in the VCS cluster. Further, after started vxconfigd, opens on the devices in error state started failing with EBUSY error. These symptoms are reminiscent of some of the issues in the past due to mode 254 not set on the HDS disk array.


Solution:

Change the settings on HDS disk array to HostModeOption MODE02: and Systems Option: 254

Please refer to page 11 of TechNote 283282: Veritas Storage Foundation and High Availability Solutions by Symantec Hardware TechNote 5.0 RP1

 

Platform	Setting 1	Setting 2	Setting 3	Setting 4
HP-UX	Minimum	Host Mode: 03	HostModeOption:	Systems Option
	Firmware:	(HP-UX)	MODE02 (Veritas	254
	50-05-22-00/00		Storage Foundation
			for Oracle RAC)