MX Records

Kevin Darcy kcd at daimlerchrysler.com
Thu Sep 21 01:00:24 UTC 2000


Eric Leslie wrote:

> Hello All:
> The company that I work for has been having some problems receiving e-mail
> recently that has been perplexing me. I am hoping that someone may have some
> ideas as to a solution.
>
> The company that I work for is hosting it's website at a virtual host
> provider (Webhosting.com), and the domain name is metzgers.com. We have an
> in-house Unix box (Cobalt Qube) that houses our mail (both Pop and SMTP). We
> are connected to the Internet by a NAT dedicated ISDN line that has 24/7
> connectivity. Static IP address is 206.165.56.209 and we have the A name
> media.metzgers.com registered to that IP address. Our router passes the
> request for POP and SMTP ports to the internal address of our Qube.
>
> The problem that I am having is that mail is delayed from some servers by 24
> hours or more. Mail coming from other locations is instant. Some of the
> messages I am receiving (people are faxing the failure reports to me) are
> Host connect failed - destination host not responding, connection reset, and
> read error.

Those are *not* DNS-related error messages. Those are caused by network or mail
server problems. DNS-related mail problems would typically show up as "Host not
found" or "Name lookup failure", or some similar wording.

> Prior to this week, the MX record pointed to pop.metzgers.com (which is what
> the webhosting company uses) instead of media.metzgers.com. Mail still got
> here, but we were having problems so we looked into the mx record and
> noticed that it was using pop.metzgers.com instead of media.metzgers.com.
> Could part of the problem be that when changing pop.metzgers.com
> (pop.metzgers.com is no longer setup to anywhere) and some servers have
> cached settings?

The TTL (time-to-live) value on the current records is 1 day. Assuming the
TTL values haven't changed, then the TTL on the old records would have been 1
day and so after 1 day the record should have disappeared from everyone's
cache. So I doubt that stale cache entries is the problem.

> Internal mail of course works fine and sending mail works fine. As I said,
> most mail seems to come through fine also, just not all of it. This is what
> stumps me. It is easier to diagnose something when it effects everything,
> but difficult when it effects part of the situation.
>
> Here is my MX record, which is maintained by Webhosting.com:
> metzgers.com preference = 10, mail exchanger = media.metzgers.com
> metzgers.com preference = 0, mail exchanger = media.metzgers.com
> metzgers.com nameserver = ns1.dynamicweb.net
> metzgers.com nameserver = ns2.dynamicweb.net
> media.metzgers.com internet address = 206.165.56.209
> ns1.dynamicweb.net internet address = 205.178.159.134
> ns2.dynamicweb.net internet address = 205.178.138.137
>
> We don't have a backup mail server as of yet, so is both preferences
> required that point to media.metzgers.com? (What is preference=0 used for?
> Webhosting.com tells me that it is the timeout value and is required,
> although as I do a lookup on other servers, it does not appear that many use
> it.)

Geez. In all my days, I've never heard *that* one before! They are clueless.
The preference value is just that: a *preference* value. Lower numbers
represent more preferred targets. They _are_ correct, however, in saying that
it is a required field. If you have only one server, just list one MX record,
and the preference value doesn't matter (I usually set it to 0 just for
readability). It looks goofy to have two MX records with the same target and
different preference values...

Tell the folks at webhosting.com to read RFC 1035 for an education on the
meanings of the various fields in the "original" DNS record types, including
MX. RFC 974 provides even more information about how MX records, in particular,
should be used in practice. Ignore all references to "WKS" records; they are
obsolete.

> Could that be causing a loop that is delaying the delivery of mail?

Nope, it shouldn't. As long as your mail server is available, the fact that you
have 2 MX records with the same target shouldn't cause any "loop" at all, since
the second MX record won't be used. Even if it is, it should be relatively
harmless for remote mailservers to retry the same mail message to the same mail
server. And your mail server wouldn't be using MX records to figure out whether
or not to accept a piece of mail either: this is configured right into the mail
server. So it wouldn't cause a loop. A loop probably wouldn't generate those
error messages, either. With sendmail, for instance, a loop usually dies with
"Too many hops" or "hop count exceeded".

I'd be scrutinizing your NAT or your mail server instead of DNS.


- Kevin




More information about the bind-users mailing list