Total Recall

Backups

This is the other most important thing, consistant and usable backups. For small installations, I typically go with tar or cpio and NFS mounted partitions. You can set up a simple cron job or script to backup all your important data to a single tape, and have it run daily or weekly out of cron.

You should also have some sort of offsite storage. This doesn't have to be with some fancy bonded company, just get it out of the same building, even if you just grab one tape every week and toss it in the glove box of your car. Really! I've been doing that for years and have never had a problem recovering stuff from the tapes.

Once you get big that you can't fit all your data on a single tape, or you get stuck with some boxes that only speak SMB, you need to get a bit fancier. For that, I've been playing with Amanda, The Advanced Maryland Automatic Network Disk Archiver. This sucker will manage tapes, do incremental backups, and can even backup those pesky Windows boxes using SAMBA.

Just one helpful hint when installing amanda, always configure it --with-debugging, because it uses a slew of undocumented files and directories, and silently fails if they don't exist.

Restores

Backing up is all well and good, but you also have to be able to restore the files as well. It's amazing how many people only try this part after a disk crash.

Restoring files is good. Recovering an entire disk is better. Being able to boot off it is ideal. Bonus points for upgrading to a bigger disk at the same time.

The typical boot procedure starts in the machine's bios, which searches for a bootable device, and, in the case of disks, typically loads the code in the first sector(s) of the disk. This secondary bootloader then performs whatever magic is required to find and load the OS.

HP-UX

make_recovery which is part of Ignite will create a tape that is bootable OS. See the man page. You can make a tape for each machine on the machine itself if they are all different.

John, We make ignite backups of our servers in addition to regular backups. We are using 10.20. Look at the man pages for make_medialif, make_sys_image, and archive_impacts. Basically you make an image of your system, create a boot file and tar the whole works to a tape. I have a procedure someplace at work for doing this. If I remember to bring it home I'll post it to you.

You have a couple of options. If the hardware on the w/stations are identical then you can use COPYUTIL (from the diagnostics -ODE- at the ISL level) that will make bit by bit copy of your disk or you can use ignite-UX's make_recovery (free at software.hp.com) and you can adjust any hardware changes. George Petrides

Or, while the machine is running.

  1. Create a new VG
  2. Add the alt disk to the VG, pvcreate the disk with the -B switch.
  3. Duplicate the LV's from the other disk (use pvdisplay here for the order of allocation) making certain that the Boot LV is first, swap LV second and root LV third in order on the device (verify this with the lvlnboot command).
  4. newfs the LV's, Boot LV should be HFS.
  5. Mount (and unmount) each of the new LV's and copy from the original (find . -depth -print | cpio -pvcmud /mount) . Do "/" lvol last and change the fstab on the target to the new vg name.
  6. Perform the lvlnboot command for each of the required LV's (boot, swap, root, and dump).
  7. Copy over you ISL Utilities with lifcp (use lifls to list them from your root disk).
  8. Reboot off that device.
Other than the reboot, all can be done while running and should take no longer than 90 minutes, much faster than COPYUTIL and it can be scripted.
Last modified: Tue Apr 6 13:40:06 EDT 1999