How can I create Volumes Groups in a IBM Storage FastT500 with HP-UX 11i_?

[Sebastian Perez]

HP-UX 11i.
Storage: IBM FastT500, multiples diks.
Conection: Fibre Channel.
I have the interface administration for the IBM Storage, and the IBM team created the LUNs.
Now, I need to create my Volume Groups and LVol, Any Suggestion_?
How Can I do_? Using SAM_? I cann't see what disk is avaible in for each LUN. I can see de device file asociated to each LUN.
I will be waiting for your feedback.
Thanks!

[Patrick Wallek]

YOu just treat the LUNs from your IBM array just like any other disks. You have the /dev/dsk/c?t?d? files, so just use those when you do pvcreate, vgcreate, vgextend, lvcreate, lvextend, etc......

There is nothing special when creating a VG.

[Michael Tully]

As Patrick has suggested, each LUN is actually seen as a logical disk. Instead of using SAM, try to see from the 'ioscan' output what LUN's you do have. If your unsure post the contents of the following command, to see we can assist further.

# ioscan -fnC disk

Regards
Michael
(not a SAM lover)

[Steven E. Protter]

If the IBM has any management console, you may need to assign the LUNS to the world wide name of the HP box.

fcsmutil /dev/td0

gives you the world wide name of the fiber card on the hp box.

Then its just ioscan and go.

There is probably an 8 lun limit on how many luns you can connect to one fiber card.

SEP

[Sebastian Perez]

Ok, thank you, and here you are the output of ioscan -fnC disk

[Sebastian Perez]

Sorry!
This version is better for the file sps.ioscandisk.txt.

[Michael Tully]

HPUX uses a virtual Bus system of numbering. There is a limit on the number of LUN's per target, but not necessarily the system.

The number may go like this:

c6t0d0
c6t0d1
c6t0d2
c6t0d3
c6t0d4
c6t0d5
c6t0d6
c6t0d7
c6t1d0
c6t1d1 and so on.

Looking at your ioscan output, there appears to be no LUN's that the system can be seen. How have you got the disk array attached to the server, directly via the fibre, or via a switch?
What is the output of:

# /opt/fcms/bin/fcmsutil /dev/td0

[Michael Tully]

That format is better.

You could now go and start using these IBM disks.

e.g. to create a new volume group etc.

# mkdir /dev/vg01
# mknod /dev/vg01/group c 64 0x010000
# pvcreate /dev/rdsk/c15t0d0 (for each one)
# vgcreate -p 128 /dev/vg01 /dev/dsk/c15t0d0
# lvcreate -n mylvol /dev/vg01
# lvextend -L /dev/vg01/mylvol

[Sebastian Perez]

Here you are the output for /dev/td0 and /dev/td1. I´m passing by the IBM Switch 3534 Model:F08 Total Storage SAN Switch.
And....well..

[Sebastian Perez]

What happend is that I want to know wich disk belong to wich LUN. Cause IBM has asociate through the World Wide Name of the FC interface, which LUN can see what server. I have tow servers hp9000 seeing the IBM Storage but diferents LUNs. That´s why I need to know wich disks belong to which LUN

[Patrick Wallek · Accepted solution]

From the HP-UX server *YOU* will not be able to determine which disks belong to which LUN. That is something that the IBM storage guys need to determine. Your HP-UX system will *ONLY* see the LUNs the the IBM storage system presents to it.

It is the IBM storage systems responsibility to present the appropriate LUNs to your machine. All you should have to provide are the WWNs for your Fibre Cards.

[Steven E. Protter]

How you assign your disks to luns is controlled at the IBM and not controllable at the hp-9000 server.

It is probably possible to set it up in whole disk mode, but if it has more than 8 disks, you may not be able to use them all.

Its important if you manage the IBM to have some idea whats going on with the disks and how they are grouped.

Its common for disk arrays to use Raid 5, but its possible to do Raid 10 as well.

If you don't run the IBM device then see the disk administrator. How the disks are laid out has a lot to do with how you plan to use them. You want Raid 10 for database work, and can tolerate Raid 5 for less intense applications.

Good Luck,

You are in good hands with Patrick.

SEP