OSEC

Neohapsis is currently accepting applications for employment. For more information, please visit our website www.neohapsis.com or email hr@neohapsis.com
 
From: Clifton Royston (cliftonr_at_lava.net)
Date: Thu Jan 02 2003 - 15:13:43 CST

  • Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

    On Thu, Jan 02, 2003 at 12:18:30PM -0800, Kenneth Maupin wrote:
    > Hi,
    >
    > We're designing a new mail system which sits behind a load balancer.
    > One component of this system is a set of hosts dedicated to queuing
    > mail for domains to which we offer secondary MX. Our intent is to use
    > the load balancer to present a single routable IP address to the
    > outside world regardless of how many hosts reside behind it. The
    > queuing hosts are configured with non-routable IP addresses visible to
    > the load balancer only. Domains for which we offer secondary MX will
    > refer to the routable address in their MX records. When the primary
    > MX host is down, mail will wait on our queuing hosts until the primary
    > comes back up.
    ...
    > We're trying to determine the correct Postfix configuration for this
    > network design. In our experiments, mail sent to domain.com arrives
    > on one of the MX queuing hosts as expected if smtp.domain.com is down,
    > but Postfix immediately tries to redeliver it to
    > mxqueue.easystreet.com because it doesn't realize that it is acting as
    > mxqueue.easystreet.com. The result is a mail loop.
     
    This was my predicted behavior, then confirmed by experiments. You are
    on the right track in your thinking and testing.

    > Our question is this: can Postfix be configured into thinking it
    > should queue mail as if it is mxqueue.easystreet.com despite the host
    > having no network interfaces configured with this name? Adding
    > mxqueue.easystreet.com to $mydestination has no apparent effect.

    SOLUTION:

    Having just been down that *exact* road, for nearly identical reasons,
    you have two options:

    * Upgrade to Postfix 2.0 and use the proxy_interfaces feature.

    * Under Postfix 1.1.x, bind the routable (virtual) IP address as a
      non-arped interface to the loopback on each of the real servers, and
      use the "inet_interfaces = all" setting in Postfix main.cf. This
      causes Postfix to recognize that the address belongs to it, and not
      try to forward mail there.

      This may be a kludge, but it works great and will solve your problems
    until you are ready to upgrade. We were in the process of
    transitioning some of our Postfix servers to a more complex virtual
    domains configuration, and chose not to confuse things with an upgrade
    at this time; we will upgrade once our new configuration is running
    stably.

      -- Clifton

    -- 
         Clifton Royston  --  LavaNet Systems Architect --  cliftonrlava.net
    

    "If you ride fast enough, the Specialist can't catch you." "What's the Specialist?" Samantha says. "The Specialist wears a hat," says the babysitter. "The hat makes noises." She doesn't say anything else. Kelly Link, _The Specialist's Hat_