hds 9900 connected to rp7400 ioscan multiple lun0

[kavanagh_1]

The rp7400 ioscan -fn of a hds 9900 is displaying entries /dev/dsk/cxt(0 to 15)d0 (lun 0) even though lun 0 has not been allocated any size or presented to the FC port and has a description of "HITACHI DISK-SUBSYSTEM". The HDS HP extension to allow 120 luns is installed. This is duplicated for every SCSI target 0 to 15. The ioscan for open-E devices is fine. (See example below). Is there any way to stop the rp7400 seeing these lun0 entries when they have not been created on the hds?


disk 6 0/10/0/0.1.64.0.0.2.0 sdisk CLAIMED DEVICE HITACHI DISK-SUBSYSTEM
/dev/dsk/c4t2d0 /dev/rdsk/c4t2d0
disk 37 0/10/0/0.1.64.0.0.2.4 sdisk CLAIMED DEVICE HITACHI OPEN-E

[Insu Kim · Accepted solution]

This is the way that HP-UX sees the HDS.
The fiber channel is emulated as F/W SCSI on the HP9000 host and the LUN number (0-119) will be translated by the host to SCSI target IDs and LUN numbers.

Bear in mind that there are basic LUN assignment rules.
1. Each SCSI target must have a LUN 0 or none of the LUNs on that target will be visible to the host.
2. Don't break a string of LUNs or the subsequent LUNs will not be visible to the host.
(Ex: if LUN 30,31,32, and 36 are assigned, the host will only see LUNs 30,31,32 but this vulnerability can be overcome by creating a pseudo device. I'll explain that below)

The array fools the host into thinking LUN 0 and the string are available by creating a psuedo device for each SCSI target ID on a given FC card, that is, a fake device will show up for SCSI IDs 0,1,2,3,4, etc even if a LUN is not assigned to that target.
This allows you to assign LUNs without having to assign to LUN 0 and also allows you to skip a LUN in the string.
That's why you see a psuedo device file on each
SCSI target ID when you issue ioscan.

Hope this helps,
Insu,

[Vincent Fleming]

You can ignore the DISK-SUBSYSTEM lines in the ioscan. They really mean nothing.

The non-consecutive LUN problem has been fixed; you need only download the patch (which I forget the number of...)

Good luck

[kavanagh_1]

Thanks for the help.

The other gotcha with the LUN0s which I have discovered is that EMS doesn't like them! The file /var/stm/data/tools/monitor/disabled_instances needs the hardware path of each LUN0 adding to it to stop EMS reporting errors on them.