From: Francis Stevens (Francis.Stevensbristow.co.uk)
Date: Wed May 10 2006 - 08:00:25 CDT

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

We had an interesting experience today which we are at a loss to
explain. It was obviously caused by a reconfiguration and is unlikely
to happen again (as we won't be repeating that reconfiguration) but I
can't explain why it happened and wondered if someone "out there" can.

We are transferring an email domain to an external service provider
(let's say domain xyz.com). Previously we have hosted this domain on
our own server, it was listed in $mydestination the MX record pointed to
us and everything was fine.

For the migration all the mailboxes on our server were duplicated on the
service provider's machine but were configured to forward all email back
to addresses on our server, the MX record was then transferred and the
local postfix configuration changed to remove xyz.com. Everything was
fine for a few days.

Today the SMTP host the the service provider drops offline for several
hours. Local user axyz.com sends an email to their neighbour bxyz.com
and is surprised when it doesn't arrive. On our mailserver the job is
in the deferred queue with a "connection timed out" error. In fact
there's lots of them :-) After a while we decide that the service
provider isn't going to fix this anytime soon so decide "Hey, no problem
let's just go back the way we were and tell our server it is host for
xyz.com again. We added xyz.com back to $mydestination and did a
"postfix reload", at this point all the deferred jobs seem to have been
"bounced" with an "unknown user" error. However any new email sent
(locally) to axyz.com is now being delivered correctly.

My first question is why did the deferred mail get bounced?

My second queustion is could I have avoided this?

Thanks in advance to anyone with the patience to read this...

FAS

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