Can't add alternate path to VG on 11.31

[Brett Simpson]

I have EMC disks with alternate paths. When I try to vgextend to add the alternate path, I get the following error:

vgextend /dev/vg02 /dev/dsk/c22t5d0
Verification of unique LVM disk id on each disk in the volume group
/dev/vg02 failed.
Following are the sets of disks having identical LVM disk id
/dev/dsk/c21t5d0 /dev/dsk/c22t5d0

Any ideas?? Thanks in advance.

[Mel Burslan]

is /dev/dsk/c21t5d0 in vg02 ? it should be.

[Zinky]

You've an 11.31 system.
Are your diskgroup uses Legacy device names?

Does "vgdisplay -v vg02" depict your member PV in legacy or agile name format?

[SoorajCleris]

Hi Brett,

I would suggest you to utilize the advantage of HPUX 11iv3. Use agile path and Native multipathing. No need to worry about alternate patch and all.

Single device file for all path and enjoy!!!!

Regards,
Sooraj

[Brett Simpson]

This what I get when I run vgdisplay:

PV Name /dev/dsk/c21t5d0
PV Status unavailable
Total PE 17262
Free PE 17262
Autoswitch On
Proactive Polling On

[Zinky]

Are these SAN Disks?

Can you post the output of:

diskinfo /dev/rdsk/c21t5d0
diskinfo /dev/rdsk/c22t5d0

also:

pvdisplay /dev/dsk/c21t5d0
pvdisplay /dev/dsk/c22t5d0

[Brett Simpson]

These are EMC SAN disks.

here are the outputs:

diskinfo /dev/rdsk/c21t5d0
SCSI describe of /dev/rdsk/c21t5d0:
vendor: EMC
product id: SYMMETRIX
type: direct access
size: 70709760 Kbytes
bytes per sector: 512

diskinfo /dev/rdsk/c22t5d0
SCSI describe of /dev/rdsk/c22t5d0:
vendor: EMC
product id: SYMMETRIX
type: direct access
size: 70709760 Kbytes
bytes per sector: 512

pvdisplay /dev/dsk/c21t5d0
--- Physical volumes ---
PV Name /dev/dsk/c21t5d0
VG Name /dev/vg02
PV Status unavailable
Allocatable yes
VGDA 2
Cur LV 0
PE Size (Mbytes) 4
Total PE 17262
Free PE 17262
Allocated PE 0
Stale PE 0
IO Timeout (Seconds) default
Autoswitch On
Proactive Polling On

pvdisplay /dev/dsk/c22t5d0
pvdisplay: Couldn't find the volume group to which
physical volume "/dev/dsk/c22t5d0" belongs.
pvdisplay: Cannot display physical volume "/dev/dsk/c22t5d0".

[Zinky]

Very strange..
I am dumbfounded.

Any chance you have EMC's SCSI "inquire" tool? With that tool (/usr/bin/inquire or /usr/sbin/inquire or /usr/local/bin/inquire) or simply "inq" - you should be able to list ech disk's serial number -- but then again the "Verification of unique LVM disk id on each disk in the volume group /dev/vg02 failed" message is bothersome.

[Bill Hassell]

What LVM is doing is reading the LVM header which has the VGID (the same for all disks in the volume group) and a PVID (unique to each disk). LVM is particularly good at keeping these numbers straight so the c22t5d0 disk is not an alternate path. The reason is that it must be the same disk using a different path. That is apparently not the case so some SAN investigation is needed.

I have attached a simple script called showLVMinfo which will report the CPUID, VGID and PVID for one or more disks:

showLVMinfo c0t6d0

Local CPU ID = 1327355523, hex value=0x4f1dd683

/dev/rdsk/c0t6d0:
CPUID=0x4f1dd683 (1327355523), VGID 0x462b913a, PVID 0x462b913a

You can use this to read hte LVM details on each disk (just like vgextend).

[Brett Simpson]

Bill,
I thought that was the whole thing behind alternate pathing. It should be the same disk but accessible via a different path. Am I wrong in my thinking? I have two HBA's going to the same allocated SAN disks.

[Zinky]

Brett,

Have you tried using the EMC inquire or inq utility? Maybe that other disk is indeed "different" on the other HBA presentation?

[Brett Simpson]

Alzhy,
I used EMC's syminq and both disks have the same serial number:

/dev/rdsk/c21t5d0 M(4) EMC SYMMETRIX 5772 02015A4008 70709760

/dev/rdsk/c22t5d0 M(4) EMC SYMMETRIX 5772 02015A4008 70709760

So, they are the same allocated disks.

Here's the paths from ioscan:
disk 51 0/0/6/1/0/4/0.108.44.19.0.5.0 sdisk CLAIMED DEVICE E
MC SYMMETRIX
/dev/dsk/c22t5d0 /dev/rdsk/c22t5d0
/dev/dsk/c22t5d0s1 /dev/rdsk/c22t5d0s1


disk 44 0/0/14/1/0/4/0.107.44.19.0.5.0 sdisk CLAIMED DEVICE E
MC SYMMETRIX
/dev/dsk/c21t5d0 /dev/rdsk/c21t5d0
/dev/dsk/c21t5d0s1 /dev/rdsk/c21t5d0s1

[likid0]

it' strange , if they have the same serial, they are the same disk.

you can allways try with agile view and vgscan.

take a look at:

#ioscan -m dsf

and check if /dev/dsk/c21t5d0 /dev/dsk/c22t5d0

have the same agile view device something like:

/dev/disk/disk4


you can also try with vgcan.

for example:

cp /etc/lvmtab /etc/lvmtab.old
vgscan -Bv -f /dev/vg02

you should get the legacy and agile view devices in you vg02, then just vgreduce legacy or agile

[Zinky]

I was just about to suggest what the pervious poster adviced. See if the agile view maps to the 2 legacy devices.

[melvyn burnard]

If you wish to use alternate paths, you need to use scsimgr to change the load balance policy from "round robin" to "simulate pvlinks"

[Zinky]

Melvyn nailed it.
I now remember 11.31 has somewhat different configuration rules under LVM.

[Zinky]

http://docs.hp.com/en/5992-4589/5992-4589.pdf

[Brett Simpson]

This is a different server, but the same thing is happening on it. Here is the output of ioscan -m dsf:

/dev/rdisk/disk54 /dev/rdsk/c21t6d0
/dev/rdisk/disk60 /dev/rdsk/c22t6d0

[Brett Simpson]

Melvyn,
How is that done? I'm searching the man pages and don't even see that as an attribute.

Thanks!

[Brett Simpson]

Ok. I found the load_bal_policy attribute, but simulate pvlinks is not one of the options. Also, scsimgr get_info shows only 1 path in the LUN info. Shouldn't it be showing 2?

[Duncan Edmonstone · Accepted solution]

Go speak to youe EMC support guys and ask them if they have the SPC-3 bit set on the FA ports on the Symmetrix that talk to the 11iv3 host.

IIRC if the SPC-3 bit isn't set, then an EMC SYmmetrix will generate a different WWID for each path to the same LUN, which causes HP-UX11iv3 to see each path as a different device.

This is *different* from the FA settings for 11iv2/11iv1, so your storage guys might not have realised.

HTH

Duncan

[Brett Simpson]

Duncan,
If I could've given you 15 points, I would've. Thanks!

[Brett Simpson]

THe only difference is that it's the SPC2 flag and OS2007 flag.

[Brett Simpson]

SPC2 flag and OS2007 flag was the problem.