postfix smtpd/mysql cf files bonked this morning

[Heitzso <heitzso@growthmodels.com>]

Please flame me quick for posting to the wrong list then
tell me where this should be posted.  Thanks.

I setup postfix/saslauthd/pam.d->mysql (for auth) and
parallel postfix->mysql (for virtual, users, aliases, etc.)
a couple of months ago on my athlon and it ran fine
(after getting it setup) for about a month.

Some three weeks ago I setup my old ss20 as my
firewall/email/dns/etc. box and moved the email
tool stack over to it.  It's worked so-so for past few
weeks with the biggest headache being the 10 second
openssl TLS setup.

Last night (5:30 am?) postfix stopped being able to access
mysql via 127.0.0.1 in it's cf files.  Result was that emails
backed up and, at its worst, smtpd was holding onto over
3,600 (three thousand six hundred) file descriptors.
cleanup was almost as bad, with over two thousand
file descriptors.  My 2.2.25 linux choked until I opened
up file-max and inode-max in /proc.  (It was after opening
up the file/inode-max that the above 3,600 and 2,000+
files open data points were picked up).

I've been crawling through the logs for an hour+,
along with 'lsof | grep smtpd | wc' type commands, etc.
to sort this out.  I fixed by changing the /etc/postfix/mysql*.cf
files hosts parameter from 127.0.0.1 over to 'localhost'.
Bingo, everything works again.  Note that I had been restarting
the whole email tool stack repeatedly without the problem
going away (and checking that all processes were flushed
with ps-aef along the way).  While that simple change
from '127.0.0.1' to 'localhost' and a 'postfix restart/reload'
fixed.

Does anyone have any idea what might have blown?
This is the third time the system choked this way,
though the first time I've crawled through the logs and
debugged.  The other two times I fixed by rebooting
the os.  I cannot promise that the other two times
file handles weren't released and linux choked were
related to smtpd/postfix/mysql.

Another data point ... 'ifconfig' showed lo and
127.0.0.1 being up.  I checked that as soon as I realized
postfix/smtpd wasn't getting into mysql via '-h 127.0.0.1'

Okay, now flame me for posting to the wrong list
and quick advise me where to post.    I just really
don't know what blew up on this sparc box and
hadn't had this experience on the athlon, though
the athlon is running 2.4.22-ac3,  both are
on 'unstable' so there's the steady thrash of packages,
and I've been trying to push key packages over to
-mv8 packages on the sparc box to get the integer math speedup
for some python/openssl/whatever packages.


Thanks
Heitzso