seriously delayed emails

[souperdad]

We are having a problem in that a lot of our emails are being delayed, sometimes up to 8 hours. As you can see from /var/log/maillog the connection is being refused by 127.0.0.1 but was in fact delivered 6 hours later (email addresses changed for privacy reasons)

Dec 12 10:25:00 mail sendmail[24332]: kBCIOsEC024332: from=<sender@sender.com>, size=103892, class=0, nrcpts=1, msgid=<4AEEB55F8E7E954897DDFAC79D82468C3CC696@gaalpsvr0281>, proto=ESMTP, relay=root@localhost
Dec 12 10:25:00 mail sendmail[24332]: kBCIOsEC024332: to=<recipient@recipient.com>, delay=00:00:03, mailer=relay, pri=133892, stat=queued
Dec 12 12:23:57 mail sm-msp-queue[31253]: kBCIOsEC024332: to=<recipient@recipient.com>, delay=01:59:00, xdelay=00:00:00, mailer=relay, pri=223892, relay=[127.0.0.1], dsn=4.0.0, stat=Deferred: Connection refused by [127.0.0.1]
Dec 12 18:13:55 mail sm-msp-queue[28917]: kBCIOsEC024332: to=<recipient@recipient.com>, delay=07:48:58, xdelay=00:00:09, mailer=relay, pri=313892, relay=[127.0.0.1] [127.0.0.1], dsn=2.0.0, stat=Sent (kBD1HrKC029435 Message accepted for delivery)

Anyone have any ideas why these emails might be sitting on the server for this time instead of being delivered?

From the scalix server I can successfully telnet to 127.0.0.1

here is the smtpd.cfg

SMTPFILTER=TRUE
RELAY accept 127.0.0.1
RELAY accept .lush.com
RELAY accept .lushcanada.com
RELAY accept .lushusa.com
RELAY Log_Reject ALL

# extra rules added to prevent open relay usage
RECIPIENT Log_Reject *@*@*
RECIPIENT Log_Reject *%*
RECIPIENT Log_Reject *!*
RECIPIENT Log_Reject *#*@*

and /etc/sysconfig/sendmail :
DAEMON=yes
QUEUE=1h

[kanderson]

How busy is your server? Sendmail might be refusing connections if it's too busy (load average of 12 or higher, from memory) Maybe a backup or something similar running?

[souperdad]

our server is very busy and quite frequently the load average is up in high teens.
The server specs are more than enough, (64bit dual core with 2 GB of RAM)

can you change the setting at which point sendmail will start refusing connections?

I'm wondering if it's a better idea to move the spam filtering to another server as according to top it always seems to be spamd that is hogging the resources

[Valerion]

With load averages that high your whole system will suffer. You HAVE to reduce that, otherwise your users will barely be able to access their email. I had this at a client's site and it affected Scalix very badly. Simply allowing sendmail to run will just make the problem worse.

Ideally you should have no more than a load average of 1 per CPU. For your system I would say anything above 5 or 6 (for short periods) or 2-4 normally would raise a red flag and cause me to investigate the matter. If you can find the culprit and move it to a different machine, it would be a much better solution.

[souperdad]

More often than not when the load average is really high, top will show spamd at the top of the list nearly every time.

[simonwelch]

We also get serious email delivery problems, this has got so bad that we instruct people to send to a gmail account. Output of uptime is detailed below and our logs don't show connection refused.

RHEL 4, Dell Dual Xeon 2850, 4 gb memory.

Simon

13:31:03 up 46 days, 18:11, 1 user, load average: 1.72, 1.55, 1.52