Date: October 4, 2004
Although pSeries servers have an excellent reliability track record, my recent experiences show no system is 100% reliable. In addition to hardware problems, you need to protect against human error (rm -r *) and software bugs.
Here are my guidelines for backing up a system
Frequency: system backups should be appropriate to the frequency of system changes, and the importance of the system. Most installations backup monthly or quarterly. You should also backup before and after any significant changes to rootvg, such as updates, and adding filesystems.
Media: tapes can become defective, so you should rotate tapes. For example, if you backup monthly, you might sequence backups over 6 tapes. If a tape fails, you would have the previous month's backup.
Also be sure to clean the tape head regularly (see manufacture's recommendations), I was recently involved with a customer who backed up their system monthly, but never cleaned the tape head. All of their backups were corrupted.
Test the restore: test the restore on a different system, by someone
other than the person who made the backup. This protects against
defective/corrupted tapes, and software bugs (see APAR IY57522).
Its also a good way to test the restore procedures. In a disaster, the
person restoring the system usually isn't the person who made the
backup. So store tapes where some else can find them.
Bruce Spencer,
baspence@us.ibm.com
October 4, 2004