discrepancies between observed disk usage and reported disk space

[Aaron Bennett]

Hello,

If I add up the reported disk space for all of the users on a partition 
and compare it to the actual observed size of the spool directory (and 
subdirectories), there's quite a large discrepancy.

For example, adding up all of the usage on one partition comes out to 
23GB, yet observed usage (du -sh spool) comes to 30GB.  This extends 
down to individual mailboxes as well, to wit:

[cyrus at calliope imap]$ /usr/lib/cyrus-imapd/quota -f user.sabrams
   Quota  % Used    Used Root
   25000       7    1800 user.sabrams

shows 1.8MB, yet:

[root at calliope user]# du -sh sabrams/
2.2M    sabrams

What I'm wondering is, is this normal?  I ran quota -f on the above user 
and it changed it from the former value of 1200 to 1800, but still not 
close to the observed usage of 2.2M.  I'm looking for a good way to 
predict actual disk usage to size our server for growth.  If this is 
just legitimate overhead, fine -- but I need to get a sense of how much 
overhead there is and where it comes from.  If there's a problem with 
our quotas, I need to fix it.

Any thoughts?

thanks,

Aaron

-- 
Aaron Bennett
Sr. Unix Systems Administrator
Clark University ITS
abennett at clarku.edu     |     508.781.7315