large packetsize stops working

["Cheryl L. Southard" <cld astro caltech edu>]

Hi,

We've got several RedHat Enterprise WS V3 Update 3 computers.  One of them
in particular seems to acquire a packetsize problem after running for
several days.  A reboot always seems to fix it.  

The first thing we notice is that e-mails larger than about 19,000 bytes
stop going through.  The end up sitting in /var/spool/clientmqueue, retrying
every hour or so and producing this error message:

Sep 27 21:50:43 mc sm-msp-queue[27357]: i8S1QPpl024368: to=cld, ctladdr=cld (231
5/150), delay=03:24:18, xdelay=00:00:00, mailer=relay, pri=328990, relay=[127.0.
0.1], dsn=4.0.0, stat=Deferred: Connection timed out with [127.0.0.1]

Another thing we notice is that we can no longer ping it from an outside
computer.  Commands like this which used to work, no longer work
	> ping -s 16500 mc
	PING mc (131.215.103.79) 16500(16528) bytes of data.

	--- mc ping statistics ---
	3 packets transmitted, 0 received, 100% packet loss, time 2012ms

Pinging with smaller packets still works:
	> ping -s 16000 mc
	PING mc (131.215.103.79) 16000(16028) bytes of data.
	16008 bytes from mc (131.215.103.79): icmp_seq=0 ttl=63 time=4.09 ms
	16008 bytes from mc (131.215.103.79): icmp_seq=1 ttl=63 time=3.73 ms

A third sympton is that we run postgresql and it usually works, but when
it gets into this weird state, we can no longer do large database queries.

Has anyone else seem this behavior?  I find it strange that only one of
our computers consistantly develops this problem.

Thanks,

Cheryl
-- 
Cheryl Southard
cld astro caltech edu