> What makes me curios is, that james is not recovering, i got
> aware of the problem 6 hours after the first connections
> problem and it did not spooled even one message.
kill -3 against the james process ought to tell us what is happening with
the various threads.
> > I am runnning James since 2004 and never got such serious problems.
There are now millions of Microsoft spambots out there. I've seen
connection rates go up by orders of magnitude. Today I was looking at 5
connections per second in a relatively quiet corner of the Internet, which
is almost as much as the Apache infratructure had to deal with in the same
time period.
> Occasionally James runs out of connections
On a day like today, happens often. Solutions are to fast fail them, tune
the TCP/IP stack, and block the worst offenders at the firewall. The
solution is not to randomly add more connections. Throwing more
resources -- more connections -- at the problem just allows more spambots to
consume them. And more connections means more RAM, more CPU, which may
spread those resources too thin to properly handle the connections that you
do have.
But JAMES should recover. I see things like:
> 27/08/06 04:05:46 ERROR smtpserver: Unknown error occurred while
processing DATA.
> javax.mail.MessagingException: Unable to retrieve the data: Read timed
out;
and
> > 27/08/06 04:05:50 ERROR smtpserver: Socket to 190.40.245.30
(190.40.245.30) closed remotely.
all the time without problem.
> Maximum number of open connections exceeded - refusing connection.
Current number of connections is 60
You're just exposing more resources to the spambots. There are infinitely
more of them than you can open connections. And you have to be careful to
make sure that you tune connections, total threads, db connections, etc., to
prevent exhausting resources.
--- Noel
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
