MX priority mechanism question

[Barry Margolin]

In article <c6oj9p$1tr1$1 at sf1.isc.org>,
 Frank Joerdens <frank at joerdens.de> wrote:

> Greetings! What I am wondering about is this: If the first MX is not
> reachable, and the SMTP daemon then queries the nameserver for the next
> exchanger (or does the SMTP daemon query both to begin with and stores
> this information?), and that turns out to be unreachable too, will the
> SMTP daemon then try the first again, or give up eventually?

It should try all the mail exchangers.  If none of them respond, the 
message should be put back in the queue, and the attempt retried a 
little later.  The retry should start again with the first mail 
exchanger.  Eventually (usually several days, and this is typically 
configurable) it will give up and return the message to the sender as 
undeliverable.

> My guess would be that it depends on the SMTP software you're using (But
> where would this be documented for e.g. Sendmail or Postfix?).
> 
> The reason I am asking is that I just put in a 2ndary MX for all of the
> domains (about a dozen) on our mail server because the router on the
> leased line that is used for our mail traffic had to be replaced.  My
> thinking being: If we run into the unexpected, and the line is down for
> longer than a few minutes, I'll change the default route on the mail
> server to our other gateway (which corresponds to a differnet DNS name,
> a dyndns.org name actually) - which is where the MX record with the
> lower priority points now.
> 
> Then I just made a potentially (can't tell because I don't understand
> the mechanism) rather evil discovery: I tried to send mail from a host
> on our network to one of the domains on our mail server, and checked the
> mail log, which kept saying that the 2ndary exchanger was not reachable.
> I don't know how this host (or SMTP daemon) came to try the 2ndary in
> the first place, because the primary was already back on line by that
> time. However, I had to start Postfix to make it try the primary again
> ... all of which makes me think now that my plan might have been a VERY
> BAD IDEA if it means that every time an SMTP daemon out there doesn't
> get the primary straight away, it'll try the 2ndary - which won't work
> since the default route points to the other gateway - and then give up
> eventually.

I'd be very surprised if postfix actually performed this way.

In any case, how is this a BIND question?  Wouldn't you have the same 
problem with any type of DNS server?  I think you should be asking your 
question in a mail-related newsgroup (e.g. comp.mail.misc if there is 
such a thing).

Also, try reading RFC 2821, which describes how SMTP servers are 
supposed to behave.

-- 
Barry Margolin, barmar at alum.mit.edu
Arlington, MA
*** PLEASE post questions in newsgroups, not directly to me ***