Loading ...
Sorry, an error occurred while loading the content.

Re: Using multiple ip addresses to prevent ratelimits

Expand Messages
  • Wietse Venema
    ... Unfortunately, you can t have instantaneous delivery to example.com when the receiver enforces rate limits. Of course my 60 second example is just an
    Message 1 of 8 , Nov 26, 2008
    • 0 Attachment
      ram:
      >
      > On Tue, 2008-11-25 at 09:21 -0500, Wietse Venema wrote:
      >
      > > ram:
      > > > Our clients set up their mail forwarding to blackberry servers
      > > > The blackberry server is doing a ratelimit and mails get held up on our
      > > > servers
      > > >
      > > > I can easily configure multiple IP addresses on the machine. Can I
      > > > configure postfix to send using different bind addresses
      > > >
      > > > I know I can change the smtp_bind_address parameter through a script but
      > > > that seems stupid having to restart postfix everytime
      > > >
      > > > Also we can never evenly spread out the mails thru different IPS
      > >
      > > There is an example in QSHAPE_README that implements delays with
      > > a non-responding destination plus smtp_fallback_relay. This might
      > > do the job for Postfix < 2.5.
      > >
      > > Postfix 2.5 has outbound rate limits per destination.
      > >
      > > http://www.postfix.org/postconf.5.html#default_destination_rate_delay
      > >
      > > You would use something like
      > >
      > > /etc/postfix/main.cf:
      > > smtp_destination_rate_delay=60
      > >
      > > Or some other delay. This delay is enforced by the queue manager.
      >
      >
      > Delay is not an option for me , because customers dont want their
      > blackberry mails delayed :-(

      Unfortunately, you can't have instantaneous delivery to example.com
      when the receiver enforces rate limits. Of course my 60 second
      example is just an example. You can reduce it to a number that is
      more appropriate.

      If you really believe that spreading out your source IP addresses
      is a solution, consider implementing it with network address
      translation.

      On the MTA host, map the client IP address+port to an address-range
      and port-range. Even if the NAT software uses sequential assignment
      (which is a bad idea afer this year's DNS debacle) it would spread
      your apparent source IP address over a range.

      If you want to choose the transport name based on the content
      or envelope attributes, use an access map or header/body_checks
      map and FILTER actions.

      /pattern1/ FILTER smtp1:
      /pattern2/ FILTER smtp2:
      ..

      Of course this delivers all mail via SMTP even if it should
      be delivered locally.

      Wietse
    Your message has been successfully submitted and would be delivered to recipients shortly.