I hate to say it, but you are in deep kimchi. If you have any chance of reinstalling the VIOS - the virtual I/O server should hold no data itself anyway - this would probably be the best and cleanest recommendation.

As you get "5.3" from the oslevel command you are on a very old (and quite surely not any more supported) VIOS-version. The current versions are based on AIX 6.1. You can see the installed version either on the command line (the command is ioslevel) or from the HMC-GUI in the LPAR-listing display. This in itself is a danger and you should update ASAP, but i have a feeling you are going to tell me that this is not possible - right?

So you are bound to try your luck repairing the rootvg. The following is described in more detail at MichaelFelts page rootvg.net:

First you need the PVID and the VGID of the missing disk:

Code:

# lqueryvg -p <disk> -vPt

You need the displayed information later to delete the disk from the VG. Next remove the dump device which has allocated some PPs on the missing disk. Create a new dump device on the available disks if necessary.

Next you can remove the disk itself:

Code:

ldeletepv -v <VGID> -p <PVID>

NOTE that even if this is "best practice" this is a surgical operation and a certain risk is involved. A productive system, though - let alone a VIOS - is not the place to try anything even remotely risky! Do yourself a favour and have current backups (preferably mksysb-images on a NIM-server) ready before you begin!

If you have the hardware to do it: the less risky way to correct this is to create a mksysb-image and edit the /image.data file before creating it (see the command mkszfile) for details). Then you restore this mksysb to some other hardware, take another mksysb and restore that to the original hardware, replacing the original. The advantage to this is that you can test, change and retest as long as you want before taking a final decision.

I hope this helps.

bakunin