Boot disk mirror

[David Owens_1]

If a boot disk is mirrored and the primary disk fails, will the system automatically reboot to the secondary (mirrored) disk? If not, why not just run a nightly cron that creates a raw disk copy of the boot disk and manually change it at failure?

Boot disks do not usually contain daily data activity anyway and this seems much easier than the overhead of Mirror Disk/UX.


Thanks,
David Owens

[curt larson_1]

it all comes down to what is the cost of your downtime.

1)If a boot disk is mirrored and the primary disk fails, will the system automatically reboot to the secondary?

no, if the mirroring is done properly, the system won't reboot, but continue running until you shutdown to replace the bad disk. This results in no unscheduled downtime.

And if you're using hot swapable disk drives, the system doesn't need to be rebooted to replace the drive either. This results in zero downtime for a primary disk failure.

2) if your primary disk fails and you reboot your system and your autoboot disk is your primary disk, no it probably won't boot. you'll need to boot manually from your secondary disk. probably with -lq flag for the secondary loader.

3) there are sites that do use your disk copy procedure to reduce downtime. others use ignite's make_net_recovery to restore systems over a network or use make_tape_recovery to restore their systems from tape.

it all depends on how much unscheduled downtime is acceptable and what your willing to pay to reduce it.

[S.K. Chan · Accepted solution]

I'm assuming your situation is you got 2 disks in vg00 (diskA and diskB) and ..
diskA=primary boot disk
diskB=mirrored boot disk
Having MirrorUX installed, you DO NOT need to do all that that raw copy stuff, the mirroring takes care of that. Two things will determine (if either one of the disk failed) whether your able to bootup or reboot unattended.
a) Setting up the boot string to escape quorum.
# mkboot -a "hpux -lq" /dev/rdsk/cXtXdX
b) Setting up your boot path
# setboot -p
==> primary
# setboot -a
==> alternate
If you got a) and b) setup and if either diskA or diskB failed your system not only continues to run but it's able to be rebooted and it'll still come up without any problem. By escaping the quorum check vg00 will be activated upon reboot regardless of how many disks are not available in the VG. When you're ready to replace the bad disk you would simply run ..
# pvcreate -B /dev/rdsk/cYtYdY
# mkboot -l /dev/rdsk/cYtYdY
# mkboot -a "hpux -lq" /dev/rdsk/cYtYdY
# vgcfgrestore -n vg00 /dev/rdsk/cYtYdY
=> recovers LVM info to disk
# vgchange -a y vg00
=> enables new root disk
# vgsync vg00
Good reference here (HP Software Recovery Handbook) ..
http://us-support3.external.hp.com/iv/data/documents/DE_SW_UX_swrec_EN_01_E/Introduction.pdf
Chapter16 and 18.

[A. Clay Stephenson]

Actually, thge truly paranoid admin will do both. Mirror/UX allows you to tolerate the failure of a disk and never miss a beat. You should never use non hot-plug disks so that disk replacements can be done without a shutdown. Disk replacement should be considered a routine task.

However, MirrorUX will NOT protect you from two things: 1) really, really bad patches 2) your own stupidity. For those things, your raw disk copy (what I call my "lifeboat" technique) has merit. I actually mirror continuously and raw disk copy (once per week during off-peak periods).

I am rapidly approaching 4 years with zero unplanned production downtime.

[Steven E. Protter]

In addition, to protect from stupidity and patches, its a really good idea to have an Ignite backup of your system.

make_tape_recovery

I have survived a disk failure of primary boot without downtime, thanks to proper Mirror/UX setup.

Since I was hot swappable, the users never noticed.

Thats a good thing too, because at my shop, they are mean.

By following the good advice earlier in this thread, you can be like me. Nobody in my building other than my department knows who I am. I'm the strange looking fellow who gets in the elevator all summer with his bike helmet on in the morning. Truly, 95% of your 1800 employees have no idea what I do.

That means I'm doing my job.

SEP