second path won't show up for a LUN

[AlexZhang]

I have a customer running HP-UX 11.31 on a rx7640 server. He presented a lun in a IBM DS5100 to both controllers and both HBA ports on the server. Each controller is zoned to a different HBA port. The server can see the lun on only one path. I told him to run "fcdutil get_remote all" from the other HBA, and it shows the controller wwn but not the lun presented to it...
I figured if the HBA can see the controller, the zoning configuration must be valid. I couldn't think of any other area I should look at.
I am going there tomorrow , so please anybody could share your idea about this?Any troubleshooting steps I should take? Thanks.
Alex

[Steven E. Protter]

Shalom,

First:
ioscan -fnC disk

insf -C disk # should not be needed but never hurts.

ioscan -m dsf
ioscan -m hwpath

If the lun is presented properly and that does not always happen, you should see it and simply vgexend to it. Skip the pvcreate step.

Post the output.

SEP

[TTr]

If you ran ioscan and you only see one path to the LUN, check one or more of the following

1. incorrect zoning
2. bad cabling
3. hardware failure: defective cable, ports, hba.
4. The DS5100 LUN presentation when creating the host and the HBAs. If one of the cables was not plugged in the HBA WWN will not show up in the DS5100. The HBA needs to be typed in manually or the host be deleted and recreated.

[AlexZhang]

customer was trying several things on multi-pathing. anyway, i found some warnings in syslog asking for "scsimgr replace-wwid".

An attempt to probe existing LUN id 0x0 faile
d with errno of 14.
Nov 17 11:05:46 vmunix: class : lunpath, instance 53
Nov 17 11:05:46 vmunix: class : lunpath, instance 10
Nov 17 11:05:46 vmunix: An attempt to probe existing LUN id 0x4050000
000000000 failed with errno of 22.
Nov 17 11:05:46 vmunix:
Nov 17 11:05:46 vmunix: Evpd inquiry page 83h/80h failed or the curre
nt page 83h/80h data do not match the previous known page 83h/80h data on LUN id
0x0 probed beneath the target path (class = tgtpath, instance = 3) The lun path
is (class = lunpath, instance 10).Run 'scsimgr replace_wwid' command to validat
Nov 16 16:46:59 sshd[22598]: SSH: Server;LType: Throughput;Remote: 10
.160.21.81-1937;IN: 91264;OUT: 10984;Duration: 1331.1;tPut_in: 68.6;tPut_out: 8.

customer agreed to use scsimgr replace_wwid on lun0 when they have a downtime...

another question, is lun0 a mandatory requirement for a HBA to see all the LUNs on a storage controller? let's say customer has lun0 having problem with a different wwid, does it mean that HBA can not see other lun1 lun2 lun3 without a working lun0?

Thanks a lot!
Alex
I attached the ioscan output as well...

[TTr]

On the IBM arrays, LUN0 is just another disk LUN, it is not used for any special purpose like in the HP arrays.

[AlexZhang]

i checked the white paper for scsimgr, here is something interests me,
"For disk arrays, if the WWID of the LUN representing the array controller (that is, LUN 0) changes, all lunpaths of that target port go to the 'NO_HW' state on a subsequent ioscan, resulting in a loss of access to the array controller. This causes the host operating system to no longer probe for LUNs behind that array controller, so the LUNs behind the controller also go to the 'NO_HW' state."
anybody can explain this in detail regarding LUN0?

[TTr]

> For disk arrays...

Which disk arrays?