finding storage at boot on luns other than 0 (sparse luns)

[Tim Mooney]

We have a variant of the SCSI "sparse lun" problem, and I would like some
advice on what the recommended method is to work around it.  Searching
for "sparse lun" and various other keywords related to it turns up lots
of hits in the search engines, but most of them are old and none of them
seem to work reliably with recent RHEL.

The systems in question are RHEL 5.2, completely up to date.  The boxes
get some storage via a pair of connections through a dual-port QLx2xxx
FC card.  The problem is that the storage appears on a lun other than 0
(there's nothing on lun 0):

$cat /proc/scsi/scsi 
Attached devices:
Host: scsi0 Channel: 00 Id: 00 Lun: 00
   Vendor: SEAGATE  Model: ST336807LC       Rev: 0C01
   Type:   Direct-Access                    ANSI  SCSI revision: 03
Host: scsi0 Channel: 00 Id: 03 Lun: 00
   Vendor: SEAGATE  Model: ST336807LC       Rev: 0C01
   Type:   Direct-Access                    ANSI  SCSI revision: 03
Host: scsi0 Channel: 00 Id: 06 Lun: 00
   Vendor: ESG-SHV  Model: SCA HSBP M29     Rev: 1.06
   Type:   Processor                        ANSI  SCSI revision: 02
Host: scsi2 Channel: 00 Id: 00 Lun: 04
   Vendor: MAYASTOR Model: lvm(58,3)        Rev: 0.9
   Type:   Direct-Access                    ANSI  SCSI revision: 03
Host: scsi3 Channel: 00 Id: 00 Lun: 04
   Vendor: MAYASTOR Model: lvm(58,4)        Rev: 0.9
   Type:   Direct-Access                    ANSI  SCSI revision: 03


The storage on scsi2, target 0, lun 4 and scsi3, target 0, lun 4 are part
of a md RAID1 mirror pair.  That md RAID1 device fails to assemble on system
startup, because the system never detects the two volumes that make up the
md device.

If we manually echo the appropriate value into the

 	/sys/class/scsi_host/hostN/scan

file, then the device shows up.  If we do it for both devices, we can then
assemble the array manually.

The question is, how do we get this to happen automatically during the
startup process, and early enough so that the devices are available for
md startup?

We tried adding

 	options scsi_mod default_dev_flags=0x240

to /etc/modprobe.conf and then rebuilding the initrd, where

 	0x200 = Scan for large LUNs
 	0x040 = Handle sparse LUN numbering

That does indeed force the system to scan for sparse luns, but the
problem is that it now scans for lun numbers > 16,384, and the fibre
targets that are providing this storage don't react well to that.  In
this case we don't need the system to scan past lun 7, but we have other
systems with this exact same problem where the storage is on e.g. lun 11,
so we need a solution that doesn't stop probing sparse luns at lun 7.

I also tried leaving out the 0x200 but instead specifying max_luns, so
we could control what the highest lun number was that the system scanned
for, but I wasn't able to make that work in coordination with
default_dev_flags=0x040 in /etc/modprobe.conf.

Am I on the right track, or is there some other supported method with
RHEL 5.x to get the system to detect sparse luns at boot time, early
enough to use those devices as part of LVM or md?

Thanks,

Tim
-- 
Tim Mooney                                             Tim.Mooney at ndsu.edu
Enterprise Computing & Infrastructure                  701-231-1076 (Voice)
Room 242-J6, IACC Building                             701-231-8541 (Fax)
North Dakota State University, Fargo, ND 58105-5164