Boot disk replacement.

[Jack Daniels_1]

Hi,

I have a situation where my boot disk DISK0 need to be replaced.

The system has booted of DISK0 which is mirrored to DISK1 in vg00.

Now that DISK0 is failing, Can i remove DISK0 from vg00 leaving only DISK1 in vg00 even though i booted of DISK0.

My plan is to remove DISK0 from vg00 and add another free disk DISK2 to vg00 and mirror it to DISK1.

I intend to replace DISK0 during our maintenance window.

I know the procedure to mirror boot disk. But
i am not sure if i can remove the disk that i booted of, from VG00 and still have the system running.

thanks.



i was wondering if i can remove DISK1 from VG00

[Uday_S_Ankolekar]

Hi,

This doc might helpful for creating a root disk.

http://us-support2.external.hp.com/cki/bin/doc.pl/sid=01a26460086dbbad44/screen=ckiDisplayDocument?docId=200000056639312

Goodluck,
-USA..

[Krishna Prasad]

I would add disk2 and mirror it before you remove disk0.

Also this will change your primary boot path if you remove disk0 from vg00.

You can change this in the gsp.

That should work although I have to admit I have never done it that way.

I have always just replaced the failed disk.
then did a

vgcfgbackup /dev/vg00
vgchange -a y /dev/vg00
vgsynce /dev/vg00 /dev/dsk/c#t#d#

[Sridhar Bhaskarla · Accepted solution]

Hi Jack,

It shouldn't be a problem. In fact, your system is running by making use of the mirror copies.

If you are replacing DISK0 with DISK2, then the process is easy.

Replace the disk and run the following commands. You would boot up by hpux -lq by interrupting the ISL.

#vgcfgrestore -n /dev/vg00 /dev/rdsk/c?t?d?
#vgchange -a y vg00
#mkboot /dev/rdsk/c?t?d?
#mkboot -a "hpux -lq (;0)/stand/vmunix" /dev/rdsk/c?t?d?
#cd /usr/sbin/diag/lif
#mkboot -vb $LIF -p ISL -p HPUX -p AUTO
(LIF is updatediaglif2 for 11.0 updatediaglif for 10.20)
#vgsync
#lvlnboot -R


If DISK2 is a seperate disk than DISK0, then you need to remove DISK0 from vg00. To reduce it, you need to remove the logical volumes on the disk from the configuration.

Do an lvdisplay -v -k /dev/vg00/lvol1 and note the pvkey (k) number. Then use lvreduce -m 0 -k /dev/vg00/lvol1 pvkey

Repeat this process for every logical volume. Each will take pretty long time if the disk is inaccessable. Once all the lvs are removed from the failed disk, you can vgreduce the disk without any problem.

Then you can use regular procedure to mirror DISK2.

-Sri

[Craig Rants]

I did this just yesterday.

I broke the mirror by making disk 1 primary. Then vgexported disk 0 out of vg00. Changed setboot to use the disk 1 hardware address. Shutdown the system replaced the disk with a new one and booted back up. Once up mirrored disk 1 to disk 0 and then broke the mirror moving disk 0 to primary, then remirrored to disk 1. Changed setboot to use hardware address of 0. Thus the server was in the original configuration. The whole thing took less than an hour.

GL,
C

[Patrick Wallek]

Is this disk a hot-swappable disk?

If so, you can do this:

vgcfgrestore -n /dev/vg00 /dev/dsk/c?t?d0

mkboot (all steps to make bootable disk)

vgsync

That is all it takes to replace a hot swappable disk. There isn't any reason to go through all the work of removing the disk from the VG and then re-extending the VG and re-mirroring.