Re: Normal CPU?

This thing won't go away.  To review, I have a Domino 6.5.4FP2 server 
running on V5R3.  Intermittently, the server CPU utilization goes through 
the roof and the server becomes unreachable to the clients, leading to 
timeout errors and duplicate messages (as they re-send messages that they 
were attempting when they timed out).

A great deal of log-hunting has found something that might be significant, 
but I don't know.  Mail from the Internet comes to us through a filtering 
server (actually, a round robin of servers).  This traffic comes to our 
Domino server via a Sonicwall firewall.  The Sonicwall has a "stateful 
packet inspection" feature that has a nasty habit of terminating 
legitimate conversations because it doesn't like something in the 
conversation.  Turning off this feature in the configuration doesn't 
actually turn it off.  Or at least not completely, I'm not sure.

Running port scans on the Domino server and on the filtering servers, we 
can see that there are incoming SMTP conversations occurring where the 
filtering server sends the data to Domino; Domino replies with "Data 
accepted"; the filtering server sends the QUIT command -- and Domino never 
receives that command, so Domino never responds.  Eventually, both servers 
give up and the whole process is repeated later according to the filtering 
server's configuration.  This problem was already occurring when we were 
on our old V5R2 iSeries and Domino 6.5.4, but the Domino server didn't 
seem to have a problem with it.  We can see in the firewall logs that it 
is dropping these packets because of the "stateful packet inspection" 
feature that has been turned off for the last several months.  There is a 
call into their tech support about this issue - again.  The reason I'm 
bringing this to the list is that  these incidents seem to coincide with 
our sky-high CPU utilizations, but I don't understand why the new server 
would be having issues with something that the old server didn't have 
problems with.

Is the dropped packet possibly a cause for what we are seeing?  I can 
imagine that both mail servers would tie up threads for an unnecessary 
amount of time, but would it cause the Domino server to go into a CPU 
frenzy?

As always, thanks for reading and sharing any thoughts.

Patrick