MTA Postfix queue build-up

[NorthernNetGuy]

Our MTA mail queue regularly builds up, with hundreds of mails in the queue. I've looked up several SKs, and have been working with TAC to help resolve this, without much progress, so i'm reaching out to the community for advice.

The error we see on all of the mails when this happens is:

(lost connection with 127.0.0.1[127.0.0.1] while receiving the initial server greeting).

running "/opt/postfix/usr/sbin/postqueue -c /opt/postfix/etc/postfix/ -f " doesnt force the mails through the queue.

The only real changes recently that could affect this blade have been moving to Jumbo take 154, which happened a little while before the issue began, and several weeks before this issue appeared we changed the main.cf to use the correct mail header.

The mails eventually release once the maximum delay time has been hit.

CPU usage is generally sitting at 50% idle, but does have occasional spikes to <10% idle.

[Wolfgang]

Did you solved your problem David ?

any information about the solution? Sounds like a similar problem we had too.

Wolfgang

[Wout_Timmermans · Accepted solution]

Hello,

I don't know if it's still relevant, but we are also having this issue.

After investigating with TAC, we were told that Postfix stops working when real memory usage reaches 90% to prevent that the memory usage gets even higher.

It is possible to increase the memory threshold for MTA by adding something to $FWDIR/conf/mail_security_config file but I wouldn't recommend it as it seems that HTTPS inspection / Application Control / URL filtering all stop working at 95% real memory usage.

The initial server greeting error you're getting in the postfix queue is basically postfix resetting all connections on the local port, which is 10025 or 10026 if I'm not mistaken.

So if your real memory usage is above 90% and MTA is enabled, you need to add more RAM to the appliance/VM or you're having a memory leak. 

Kind regards.

Wout