Vacuum problems

Hi everyone...

I have a database that is currently about 25 gigs on my primary DB
server running Postgres 8.2.9, and two others that are less than 1 gig
apiece. The DB server is a quad proc, quad core, 64 gigs of memory, 5
drive RAID5 array, so it has plenty of horsepower. Until about three
weeks ago I was running a nightly vacuum analyze and a vacuum full
analyze once per week.

This is what I was running for the vacuum full command:

vacuumdb -a -e -f -z -v -U postgres

The nightly vacuums have been working flawlessly, but about three weeks
ago the vacuum full started failing. It was taking about 5-10 minutes
normally, but all of a sudden it started hitting the command timeout
that I have set, which is at 60 minutes. I thought that it may be a
corrupt table or a large amount of content had been deleted from a
database, so I built a script to loop through each database and run a
vacuum full analyze on each table individually thinking I would find my
problem table. The script finished in 5 minutes!

This is what I'm running on each table now in my script:

vacuumdb -d $DB -t $TABLE -e -f -z -v -U postgres

As I understand it, the "vacuumdb -a" command basically does the same
thing as my script. So why is it timing out while my script finishes in
5 minutes or less? Is the "vacuumdb -a" command doing something that
I'm not?

Now that the holiday season is past I will be upgrading to the latest
8.2 release as soon as possible, but I need to get this figured out
first unless it's a bug that the upgrade fixes.

I'd appreciate any advice or explanations you guys can send my way.

Thanks,

Scot Kreienkamp

La-Z-Boy Inc.

skreien(at)la-z-boy(dot)com

734-242-1444 ext 6379