Andy,
If you have messages hanging around in your spool, remove all of the
messages from your spool and then restart and see if it doesn't creep up
like it did before. If that stops it, then bring in pairs of messages a
few at a time to see if something triggers the behavior. When found,
test it again to confirm and then report to Ipswitch.
It would generally be Queue Manager that barfs on formating. If you
find a culprit and something unique to that culpruit, then you can block
it with Declude before it hits Queue Manager. Generally it is zombie
spam that will trigger exceptions, and these patterns can clear up as
soon as they change campaigns. Having a greylisting gateway in front of
IMail such as Alligate will remove most of the zombie spam without
creating false positives, and help to make one's server more stable.
Matt
Andy Schmidt wrote:
Sorry for cross-posting. I'm not sure whether Declude and/or Sniffer
still rely on the Paged Pool -- and whether their usage would be
reported under the Imail QueueMgr.exe or under some other .exes? So I
have 3 possible culprits...
The symptom started as a Webmail problem because customers noticed
they couldn't send emails any longer due to "Bad Socket State".
However, when I log into the physical machine, the REAL problem is
that I cannot open ANY TCP/IP connections to any IP address (on that
same machine or on neighboring machines). I can still PING (is ICMP
works), but TELNET, FTP, HTTP - all are unable to create a socket.
FTP.exe <ftp://ftp.exe> reported that it doesn't have enough buffer
space.
That caused me to turn on "Task Manager" and add the columns for "VM
Size" and "Paged Pool". Normally, the various processes only use less
than 100K of "Paged Pool" - even the IIS Web Process uses only 300K.
However, QueueManager was up to 4500K. Restarting QueueMgr.exe
service reset it to 200K or so. But, I there are time spans where it
consumes an extra K every second - now already up to 800 K again -
before it levels off for a while and then keeps doing it again.
Oddly enough, this problem only started yesterday - even though
2006.21 has been running since 7/16/2007 - and now seems to accelerate
(happened twice today!)
My obvious suspicion is that there is a 'certain' email or type of
spam that's causing this QueueMgr behavior - what else would account
for this to start happening NOW.
