Postfix qmgr process causes heavy overload on mailservers

Question

We are using Postfix as MTA for our e-mailmarketing software and once in a while we see that the load on one of the mailservers rises above 5.

The load is caused by the qmgr-process which is the heart of Postfix and I see that it is consuming a lot of CPU resources. The process seems to be stuck because after 15 minutes it is still doing the samething and still increasing the load. Once I restart the postfix service the load rapidly decreases to below 1 and Postfix continues to send e-mails without any problems.

I'm wondering if anyone else has encountered this problem and if people have suggestions on how to prevent it.

The problem shows up on all our mailservers but almost never at more than 1 at the time. It seems to be triggered only when we are sending a mailing but the size (10 or 100.000 e-mails doesn't seem to make a difference). It maybe happens once a week or even less often and the time and day is also different every time. We tried to solve the problem by decreasing the amount of messages qmgr is allowed to process but this didn't solve it.

We are using Postfix 2.5.5 on Debian Lenny 5.0.8 (postfix is installed through the default Debian repository). No special messages can be found in the logs (syslog, messages, mail.*).

Thank you for your time

Answer 1

There is a thread in the postfix users mailing list about a problem with the same symptom.

http://www.mail-archive.com/[email protected]/msg08961.html:

'The issue found applies only to "rate-limited" transports [...] The original 2.5.x code is correct for "oqmgr", but not for "qmgr" '

http://www.mail-archive.com/[email protected]/msg09030.html:

"I will release this (correcting patch) as part of Postfix 2.5.7.

Meanwhile, you can use oqmgr and it it will an all likelihood perform just as well"

Hope it helps.