Prev: Allow specific IP to send mail without username/password
Next: Outgoing Approval Queue - Yes This is a Dumb Idea
From: Erik Logtenberg on 26 Apr 2010 07:39 The Doctor wrote: > Question has anyone in postfix and / or amavis ever seen > repeating e-amils with binary attachments to them? I think many subscribers to this list have seen mail loops of all sorts, perhaps even purposely created such loops (for instance to test throttling features). So the direct answer to your question is without a doubt: "yes". If you have a specific problem that you'd like to solve, please elaborate a bit on the situation, include relevant parts of logfiles and/or headers and most importantly describe exactly what happens. I.e. are your duplicated mails delivered locally, or sent out by SMTP, are the mails exactly the same or are certain properties different (queue ID's for instance). Does the duplication only happen in certain cases - you mention binary attachments, are mails without attachments or with other types/sizes of attachments delivered without problems? Erik
From: Wietse Venema on 26 Apr 2010 08:29 The Doctor: > Question has anyone in postfix and / or amavis ever seen > repeating e-amils with binary attachments to them? You have an SMTP-based filter that takes two minutes to process a 5MB message. See my logfile analysis of a few days ago. This causes the sender (into the filter) to time out. As per the SMTP protocol, this is a "temporary" error condition. The SMTP protocol requires that senders retry delivery after a "temporary" error, until the message expires in the queue. Wietse
From: Charles Gregory on 26 Apr 2010 09:17
On Mon, 26 Apr 2010, Wietse Venema wrote: > You have an SMTP-based filter that takes two minutes to process a > 5MB message. See my logfile analysis of a few days ago. > This causes the sender (into the filter) to time out. As per the If it helps any, I recently ran into a sudden increase in filtering time (and the resulting repeating mails) as a result of a CLAMAV update. I discovered that my script was using 'clamscan' rather than 'clamdscan' (with the 'd' to use the clamd daemon). I started using clamdscan and processing time dropped back down to the appropriate 10-20 seconds. - Charles |