distinguish physical disks or LUNs

[Ming Zhao]

hello,

"ioscan -fnkCdisk" will tell you all disks or LUNs. But you know some of paths (primary and alternate path) are pointed to the same disk or LUN. So I need to know which paths are pointed to the same disk or LUN. Although "xpinfo", "syminq", or other 3nd vendor's commands can provide useful information, but they can't provide local disks or other vendor's disks and paths information. Any idea is welcome.

Thanks in advance!

Ming

[Rajeev Shukla]

Hi,
How is the XP connected to the systems. Is it via hub or a switch.
You'll have to closly see the hardware path the LUN No. will be same for a single disk seen from 2 paths.(the LUN ID is uique) Take the ioscan output and compare only the zone number (if connected through same switch but 2 controller on XP). And the controller number will be different if they are connected from different fiber channel HBA on server.

Let me know how you have your connection to the server from XP in which case i might be able to explain more clearly.

Rajeev

[Michael Elleby III_1]

Hello-

Most NAS devices (such as the Symmetrix), are configured to use the same naming convention on the device files when it is connected to an HP server.

Maybe you already know this, but the easiest way for me to begin to identify the device(s), especially disks that are configured with multiple luns (i.e. coupled with primary and alternate paths pointing to the same disk) is to review the output of STM (Support Tools Manager) that can tell you what disk controllers exist and what devices to look for on that controller, and then of course to review my device file info from the ioscan and check the d value in the device file name which would indicate the presence of logical disks (luns) on one disk.

Just my $.02 worth-

Mike-

[Thamm]

Hi.
Use the following way:
the comand xd -j 8200 -N 16 /dev/rdsk/cXtYdz
will give you VG information.
Eg:
xd -j 8200 -N 16 /dev/rdsk/c6t9d1
0000000 12ba 7434 3e00 6fcd 12ba 7434 3b34 406e
0000010

If one disk will belong to one VG and you execute that comand for another disk, you will be able to find the alternate link: The alternate link has the same VG Information.
Eg:
xd -j 8200 -N 16 /dev/rdsk/c5t9d1
0000000 12ba 7434 3e00 6fcd 12ba 7434 3b34 406e
0000010
===> Alternate Link
xd -j 8200 -N 16 /dev/rdsk/c5t8d0
0000000 12ba 7434 3e00 6fcc 12ba 7434 3b34 406e
0000010
===> Not alternate Link!


xd -j 8200 -N 16 /dev/rdsk/c9t6d0
0000000 12ba 7434 3e01 78bf 0 0 0 0
0000010
===> Does not belong to any VG

[Armin Feller]

Hi Ming,

if you are working with a HP supported disk array (EMC, XP, NIKE, ...) and patch PHCO_26382 is installed then you can see in SAM all disks, LUNs ones without alternate paths shown as divice.

SAM scans all the devices files and works out which are duplicate paths. You will find at the point NUMBER OF PATH a 2,3,4 for each disk, LUN which are reachable by a alternate link.

Sometimes SAM is not as worst as you are thinking about ;-)

Best regards,
Armin

[Jean-Louis Phelix]

Hi,

Thamm is right, but I could had some more info :

* use xd -tu -j8200 -N16 /dev/rdsk/xxx

* -tu option will give you 4 numbers :
0000000 2005332703 1026912898 2005332703 1026912900
0000010

. first number is the cpu id where pvcreate was done (identical to "uname -i" on this system). Sometimes useful to find out which system has overwritten one of my shared disks ...

. second number is the timestamp when the pvcreate was done. Display in ascii using for example :

#include
main (argc, argv)
int argc;
char *argv[];
{
time_t t;
t=atol(argv[1]);
printf("Timestamp is %s", asctime(localtime(&t)));
}

In my case it would display :

Timestamp is Thu Feb 21 10:39:33 2002

. third number is same as first but for vgcreate command

. fourth number is also same as second for vgcreate

* So for alternate links, all four fields should be identical, while for 2 PV of the same VG only last 2 fields should match.

Obviously it works only for LVM disks. So before using LVM, it's harder. Last, I think that only vgreduce resets last 2 fields to 0. I dont think that vgexport would do it, but it's quite easy to check.

Regards.