I have a Linux CENTOS Server running PostFix. It has no mailboxes, but is a mail gateway, for several Domains in another destination Server, which is also a Linux CENTOS running Sendmail. So, after checking the emails with antivirus and antispam, the Postfix Server delivers them to the Sendmail Server which hosts the real mailboxes.

Many of our customers are claiming they don't receive several emails, or receive them with many hours delay, so we monitored the /var/log/maillog file at the Postfix Server, and found out that those emails are not being delivered to the Sendmail Server and are deferred with messages like:

Aug 23 11:48:58 srv7 postfix/smtp[618]: 980C773D64B: to=, relay=srv6.multisitesdominios.com.br[], delay=2375, status=deferred (conversation with srv6.multisitesdominios.com.br[] timed out while sending message body)

Most of these emails have attachments, so probably they are a little large. We realized that small emails pass and are received normally.

My questions:

  1. What is the real reason of this problem, is it really the emails size?
  2. Is there any Postfix parameter we should tune to avoid it?
  3. Would the problem be at the destination Server (the Sendmail one) and not at this "gateway" Server (the Postfix one)?
  4. What would be the definite solution?
  1. It's indirectly the size. It takes too long without receiving progress on the data so that either side hangs up
  2. smtp_data_xfer_timeout. It defaults to 180 seconds which should be OK in every scenario.
  3. It can be on both. But probably it is the network in between.
  4. Check why the communication stalls.
| improve this answer | |
  • Perhaps all of the available programs to handle SMTP are saturated, and the next connection from Postfix can't get a response from Sendmail because Sendmail is busy? It's been years since I tried to crawl through sendmail.cf, but I do know postfix controls how many smtp connects you can get going with master.cf. – Avery Payne Dec 3 '11 at 2:24
  • 2
    @AveryPayne: No, that can't be it, because the error is timed out while sending message body; this means it got at least as far as sending the DATA command (and probably sent part of the body). If the receiving server didn't have capacity to handle additional transactions, it would have timed out before HELO (and probably before the SMTP banner). – Flimzy Dec 3 '11 at 18:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.