Boot disk LVM panic after changing SCSI ID

[Timo J]

Initial configuration:

One disk, /dev/dsk/c0t3d0 with 11.11 installed with lvm.

Next I changed the SCSI ID of the disk to 2, so new disk device is /dev/dsk/c0t2d0. And that's when the problems started. When I reboot the system, LVM Panic occurs and I'm back at the ISL prompt.

So I searched through the forums and tried the procedures I found. First I booted from IPL to lvm maintenance mode. Next I exported vg00 away, created /dev/vg00/group again (mknod /dev/vg00/group c 64 0x000000)and used vgimport to read the lvm structures from disk to vg00 (vgimport /dev/vg00 /dev/dsk/c0t2d0). After that, 'init 3' and everything looked ok; /etc/lvmtab OK, vgdisplay vg00 showing that vg00 contains that /dev/dsk/c0t3d0. But, next reboot the same LVM Panic occurs.

[Peter Godron]

Mikko,
can you please let us know what you were trying to do?
You adding a new data disk (SCSI ID 2).
You were not trying to boot of this blank new disk?
Could you not boot to your first disk and then configure (pvcreate/mkfs ...) the new disk?

[Timo J]

I thought I made my point clear ;|

I'll try to explain a little more.

Everything was working ok after the initial installation, where vg00 contained /dev/dsk/c0t3d0.

But then I changed the SCSI ID (jumper on disk) on that disk to 2, so the new disk device is now /dev/dsk/c0t2d0. And after that when rebooting, ISL Panics, because it can't find /dev/dsk/c0t3d0 anymore. So where to tell the system that "Hey, boot form /dev/dsk/c0t2d0".

When checking the primary boot path from IPL, it is 10/0.2.0, which is OK. When checking the primary boot path with command 'setboot' from OS, it is 10/0.2.0, which is ok.

Following is the error message when trying to boot from primary boot path:

"LVM : Activation of root volume group failed. Either no physical volumes are attached or no valid VGDAS were found on physical volumes."

[Torsten. · Accepted solution]

Don't boot directly from LVM maintenance mode - reset the system!

You need to install the new device files for the disk (insf -e) and re-create the the following information:

# lvlnboot -r /dev/vg00/lvol3
# lvlnboot -b /dev/vg00/lvol1
# lvlnboot -s /dev/vg00/lvol2
# lvlnboot -d /dev/vg00/lvol2

Why you want to change (and lower) the ID of this disk?

[Kent Ostby]

Prior to the lvlnboot commands that the previous poster lists, you should type:

lvrmboot -r /dev/vg00

[Timo J]

Thanks for quick responses.

Torsten, about your question the answer is: 50% curiosity & 50% real need for information. I'm going to do some updates and disk changes with pre-installed disks on production systems soon, and I don't want to find out at middle of the night that my system won't boot anymore just for SCSI ID changes. That's theoretical; easiest way of course is to keep the ID:s as they are.

[Ajitkumar Rane]

Mikko,

Is there a device file present on the system for c0t2d0? as you said after changing the scsi id to 2 the systems panics so it hasn't reached the point to run the ioinitrc scritp to creat the device file for this new device. But since you could import the VG using the the new device file its not a problem. You need to run the lvlnboot as described in one of the threads above. Also one thing to notice is you say the vgdisplay still shows c0t3d0.


rgds

ajit

[Timo J]

Ajitkumar: Yes, that was a typo on my first post, I meant c0t3d0.

Yes, there was the right device file.

At some point I managed the situation by editing /stand/bootconf to contain the right disk device. After that system booted ok. Unfortunately I'm not sure if I did those lvlnboot operations before it ;|

But to be sure I did some testing again and Torsten's advice was perfect.

[Josiah Henline]

The steps needed after changing the SCSI ID to 3:

Make sure the disk device is power cycled so the new SCSI ID takes affect.

From the BCH type "sea IPL" and make sure the drive drive shows at address 3.

Boot the system to LVM maintenace mode:
bo NEWDISK_HARDWARE_PATH #from the BCH
Y #to interact with the IPL
hpux -lm #from the ISL prompt

Once you are in LVM maintenace mode:
/sbin/vgexport vg00
/sbin/mkdir /dev/vg00
/sbin/mknod /dev/vg00/group c 64 0x000000
/sbin/insf -e
/sbin/iosocan -fnCdisk #take note of the new device name of the disk. It should be /dev/dsk/c0t3d0
/sbin/vgimport /dev/vg00 /dev/dsk/c0t3d0
/sbin/vgchange -a y /dev/vg00
/sbin/mount /
/sbin/mount /sbin
/sbin/mount /var
/sbin/mount /usr
/sbin/lvlnboot -b /dev/vg00/lvol1
/sbin/lvlnboot -s /dev/vg00/lvol2
/sbin/lvlnboot -d /dev/vg00/lvol2
/sbin/lvlnboot -r /dev/vg00/lvol3
/usr/sbin/setboot -p NEW_HARDWARE_PATH
/usr/sbin/setboot #Verify the output
/etc/reboot -r #Running an "init 3" from LVM maintenace mode may cause some unwanted side affects. Rebooting the server will verify the new configuration.