Re: ASM on AIX and Virtual I/O (VIO)

From: <erikw_nl_at_yahoo.com>
Date: 10 Jul 2006 00:47:20 -0700
Message-ID: <1152517640.448797.41670@h48g2000cwc.googlegroups.com>

Kai Nielsen schreef:

> Hi!
>

Thx very much for thius information. The PVID/ASM relationship contains some surprises.
> * erikw_nl_at_yahoo.com <erikw_nl_at_yahoo.com>:
> > Does anbody have experience with providing disk storage through a
> > Virtual I/O server to ASM ?.
>
> Yes.
> No problems. At least no problems related to ASM on VIO disks.
>
> > Can Virtual I/O server handle disks without PVID ?
>
> Yes, why shouldn't it?
>

A software technician of IBM warned that ASM and VIO won't work together because of non-existent PVID info on ASM disks. I did not have any formal confirmation of this until now.

> aix100t001:/root#lspv
> hdisk1 00cdfb0d2c230041 orapgmvg active
> hdisk2 00cdfb0d2c507815 rootvg active
> hdisk3 00cdfb0d2c231518 orapgmvg active
> hdisk8 00cdfb0d2c230c8c None
> hdisk9 00cdfb0d2c232da6 None
>
> So actually there are PVIDS. The disks not in any VG are the ASM disks.
And here is the catch of ASM. The PVID is stored on the first block of the disk:
For instance:
alligator# lspv | grep hdiskpower0

hdiskpower0     00c7a3bbfc2b051b                    systemvg

active

alligator# lquerypv -h /dev/hdiskpower0

00000000   C9C2D4C1 00000000 00000000 00000000  |................|

.....
00000080 00C7A3BB FC2B051B 00000000 00000000 |.....+..........| ....
An ASM disk looks as follows:

hdiskpower17    ASM                                 None

alligator# lquerypv -h /u02/oradata/dev//hdiskpower17

00000000   00820101 00000000 80000000 F6B9A122  |..............."|
00000010   00000000 00000000 00000000 00000000  |................|
00000020   4F52434C 4449534B 00000000 00000000  |ORCLDISK........|
00000030   00000000 00000000 00000000 00000000  |................|
00000040   0A100000 00000103 44474143 49315F55  |........DGACI1_U|
00000050   4E315F30 30303000 00000000 00000000  |N1_0000.........|
00000060   00000000 00000000 44474143 49315F55  |........DGACI1_U|
00000070   4E310000 00000000 00000000 00000000  |N1..............|
00000080   00000000 00000000 44474143 49315F55  |........DGACI1_U|
00000090   4E315F30 30303000 00000000 00000000  |N1_0000.........|
000000A0   00000000 00000000 00000000 00000000  |................|
000000B0   00000000 00000000 00000000 00000000  |................|
000000C0   00000000 00000000 01F5964B 1926F000  |...........K.&..|
000000D0   01F59BB2 3D839000 02001000 00100000  |....=...........|
000000E0   0001BC80 00009000 00000002 00000001  |................|
000000F0   00000002 00000002 00000000 00000000  |................|

As you can see, on location 00000080 the PVID is set to zero's by ASM.

The AIX system still thinks it has a PVID because initially it stores the PVID in its ODM, and keeps it there until somebody does rmdev on that disk, then later you will see it does not have PVID anymore. If you then do chdev -l hdiskpower17 -a pv=yes, you destroy the database ;-) (thx mr. Ellison).

The fact that your system is working fine, is a good sign. I will ask IBM for a confirmation on this.

Thx for all the info.

Regards,

Erik Received on Mon Jul 10 2006 - 02:47:20 CDT