On 1/25/2015 3:47 AM, Chris Robinson wrote: > > I use an Amazon SES relay that has a max message size of 10MB. I > have a small client base who sometimes send large documents as > attachments. Sometimes they will send to multiple recipients which > may include external and internal (local) addresses. > > When this happens postfwd will get a hit on the EOM rule that > detects size>9MB. The rule's action is "FILTER smtp:" in order to > bypass the relay. This rule also has a pcre item that is meant to > exclude local recipients from the hit but which was ignored (I > realise now it's because we are dealing with the whole message) with > the result that the external recipients got "mail loops back to > myself". > > I thought I might be able to use a before-message filter because the > README says "1. Re-inject the mail back into Postfix via SMTP, > perhaps after changing its content and/or destination." but I was > not clear on how to change the destination and, in any case it would > still apply to all the recipients I assume. > > I thought I could use a postconf concurrency parameter to force > postfix to split all messages into single recipient messages right > at the SMTPD beginning so that postfwd got only single recipient > messages but I could not make this happen. This would be the ideal > solution. > > I got these from Noel Jones and Victor Dukhovni: > >> Noel Jones wrote: >> > >>> On 1/23/2015 10:41 AM, Viktor Dukhovni wrote: >>> For custom per-user routing that depends on policy-based factors >>> force an extra SMTP hop for the entire message via a transport with >>> recipient concurrency set to 1. On the receiving end of that you >>> can do per-user FILTER actions as messages will have only one user >>> at a time (delivered as many times as there are recipients). > >> For this specific application, it should be sufficient to use a >> policy service FILTER action to send only big messages to a second >> postfix instance, then let that second instance deliver mail >> "normally" (normal local delivery, empty relayhost setting). There >> should be no need to set the concurrency to 1. > >> Messages under the target size can be delivered by the primary >> instance to the relayhost, same as they currently are. > >> One way adds a second hop for all mail, the other makes some mail >> "special". I don't really like either of these choices. If it's OK >> to send big mail directly, why not just send it all directly. >> >> -- Noel Jones > > Victor's solution doesn't mention a second instance of postfix but I > take it that's implied(?). Presumably Noel's solution is the same as > Victor's without the concurrency issue and they both require me to > tackle multi-instance postfix. It's additional learning curve I had > hoped to avoid. > > Noel, the reason I don't send mail out directly is that a managed > relay handles all the hassle of outbound mail filtering and I never > have to worry about my IP being blacklisted again because of some > slip up or oversight. And with my low volumes the cost is trivial.
You've somewhat painted yourself into a corner here. You either need to deliver all mail directly, or use multiple instances (details of the multi instances notwithstanding), or use something other than postfix. I still think direct delivery is your best choice. Good luck. -- Noel Jones > > Wietse, unless you can come up with a lovely simple way of handling > multi-recipient routing, or unless Victor's solution does not > involve a second instance of postfix, I think I will have to follow > Noel's advice. > > Thanks all, > Chris Robinson
