HP-UX 10.20: Full System Recovery from fbackup

[Scott Corzine]

Hello-

I'm trying to recover a D370 w/ LVM from a complete fbackup tape (created in single-user mode and verified). So far I've come close, but not quite suceeded (kernel panics during boot).

I'm now aware (from this forum's archives) that I should have downloaded Ignite and used make_tape_recovery, but it's a little too late for that...

Being somewhat new to HP-UX (but experienced with other unixen), I foolishly read the "HP-UX System Administration Tasks" PDF which indicates that fbackup is the best choice for backup (and doesn't mention make_recovery). I've been through several other manuals and found that they have conflicting instructions (do/don't use COPYUTIL, etc), but even the "Expert System Recovery" section didn't cover this.

I tried booting off of the Installation CD and using the recovery options, but found that the RAM FS was simply too small to load the needed programs. But I was able to get the LIF in a good state.

Based on advice, I tried doing a minimal install with the same lv sizes for vg00, then adding the additional vg/lv's, and running frecover in single user mode over the running system. Unsurprisingly, that didn't boot, but using the recovery mode and maintenance mode (hpux -lm) I've been able to get it to the point where the kernel loads and panics.

Sigh. Is there a way to perform a full system recovery without a make_tape_recovery? And how does one deal with a system w/o a tape drive when there's no Ignite boot server?

Thanks,
-Scott-

[Rajeev Shukla]

The method you are trying should also work provided when you took the backup all the root volume group filesystems were mounted and the same when you recover. Because by default if /usr, /var etc..are separate filesystem they do not get mounted in single usermode. You have to manually mount them and then take a full fbackup.
Can you load the tape on some othe system and see if it has got the full list of files, specially latest kernal in /stand which you had prior to system going bad.

Rajeev

[A. Clay Stephenson]

THe simple answer is no; you really need to use Ignite. THe most fundamental problem is building and sizing the vg00 LVOL's and making filesystems and mounting them. THe recovery media is really intended for you to fix a mildly sick box - repair the / or /stand filesystems, e.g. You really need either a local tape drive or an Ignite server to recover from this situation. The good news is that Ignite works amazingly well so learn to play by the rules and use it.

[Armin Feller · Accepted solution]

Steps to Prepare for Recovery
-----------------------------

This is a list of things you can do ahead of time that will make
recovery easier after a failure:

1. Regularly print the output of the following commands:

/usr/sbin/ioscan -fk
/usr/sbin/vgdisplay -v
/usr/sbin/lvlnboot -v
/usr/sbin/lvdisplay -v /dev/vgXX/lvYY (for every logical volume)
/usr/sbin/pvdisplay -v /dev/dsk/c#t#d0 (for every LVM disk)
cat /etc/fstab
fax = cat /etc/fstab

2. Make sure you have a reliable full system backup.
Use the SAM Backup and Recovery, fbackup, or any other supported
method. It may be useful to use COPYUTIL after each major
system change. This is done from the Support Media.

Note: This will obviate the need for step 2 below in the following
section.

3. Make an extra copy of these files in a non-root volume group:

/etc/passwd
/etc/group
/etc/lvmtab
/etc/lvmconf/ (entire directory)
/etc/fstab
/etc/rc.config.d/ (entire directory)


Steps for the Recovery
----------------------

1. If possible, perform a vgexport(1M) of your other volume groups,
except for vg00. Make sure that you use the option for map files.

Example: vgexport -v -m vg00.map /dev/vg01

2. Reinstall on your root disk(s), Specify all of the logical
volumes and disks needed for vg00. Make sure to remember that the
device files for these disks may remap. Expand any file systems
as needed at this time.

3. Restore all files in the root vg (usually vg00) from backup,
except for the files is this list:

/etc/fstab
/etc/lvmtab
/etc/lvmconf/
/etc/ioconfig
/stand/ioconfig
/dev/dsk/
/dev/rdsk/
/dev/rmt/
/dev/vg*
All non vg00 mount points

4. Restore the mapfiles from tape. If you don't have map file skip
this step.

5. Perform a vgimport of the other volume groups, specifying the
map files if applicable. Remember to specify all the disk device
files all at one time.

Example: vgimport -v -m vg02.map /dev/vg02 /dev/dsk/c1t4d0
/dev/dsk/c1t3d0

6. Edit /etc/fstab to reflect all file systems properly.

7. Refer to the information collected in Step 1 for any
additional "restoration" of the system configuration.

8. If you didn't have a map file, then rename the /dev/vgxx/lvoly
and rlvoly filenames to their original names (for example,
/dev/vg01/foo and rfoo).

[Scott Corzine]

Armin,

Thanks, that's the answer.

I'd muddled through much of it beforehand, but hadn't adequately handled the /dev links, leaving two different lvol's on the same minor device number.

Everything worked fine (after I fixed that, scraped a little more cruft off, and rebuilt the kernel for good measure).

Thanks again,
-Scott-

(And thanks to Rajeev too -- that's (almost) exactly how I made the backups originally)