Even properly configured, seeing intermittent "MX list for foo.com. points back to mx4.mailhost.com"

From: loonux fc (loonux_at_gmail.com)
Date: 03/10/05

  • Next message: Joe Reid: "SunBlade 1000 diag problems - rephrase"
    Date: Wed, 9 Mar 2005 19:51:11 -0500
    To: sunmanagers@sunmanagers.org


    I am trying to debug a mysterious problem. We run several mail
    servers, of which in total we probably handle upwards of 1 million
    messages a day. Our mail hosts are set to relay mail to their
    ultimate destination. In order to do that, sendmail is configured as

    1. "foo.com" is added to /etc/mail/relay-domains
    2. The following entry is added to mailertable:

      foo.com smtp:[mailhost.foo.com]

    3. An entry is made to /etc/hosts for mailhost.foo.com:

    This is how it should be, and it works 99.99999% of the time.
    However, on what appears to be completely random occasions with
    completely random customers, the dreaded message will pop up:

    Mar 9 08:09:38 mx4 sendmail[7284]: [ID 801593 mail.crit]
    j29G8qR07010: SYSERR(root): MX list for foo.com. points back to

    And, the message is bounced back to the sender with the usual "local
    configuration error" line. However, all around this line are signs
    that the system is indeed working as it should; for some reason, it
    simply appears that sendmail is not honouring the mailertable entry,
    or is somehow unable to consult /etc/hosts and is ignoring the
    mailertable entry. There appears to be no correlation with anything
    bad happening on the system: there are no other errors reported, and
    other mail transactions happening at the same time are processed
    without error.

    I'm running Solaris 8/sparc on Netra T1's with 256Mb memory, running
    sun-supplied sendmail (8.11.7p1+Sun). My /etc/resolv.conf lists
    "files dns", although I do not think this is applicable in this case.
    The only odd thing about our mail handling is that we run two daemons:
    one to accept inbound mail and drop it into a queue, and another to
    flush that queue on a regular interval.

    Note that A records for "mailhost.foo.com" do not exist in DNS, only
    in the system's /etc/hosts.

    Has anyone else experienced such strange intermittent issues as this?

    Thank you for any help that you can provide.
    sunmanagers mailing list

  • Next message: Joe Reid: "SunBlade 1000 diag problems - rephrase"