Reply to address

Hi,

Having a problem with a few non local LAN individuals replying to
tickets. For some reason the reply to address for some users is
rt@machinename instead of rt@domainname. I have the proper reply to
address listed in Site_Config and even have canonical mappings setup in
postfix. I have the same problem on multiple queues.

What is puzzling is that for most users local or not the reply to
address is either the default in Site_Config or the configured queue
address and works as expected.

Has anyone else seen this?

RT 3.0.7_1
Perl 5.8

Don’t know about postfix, but I ran into something similar
recently. I configured a machine whose dns arecord was
"machinename" and a cname of “domainname” pointing to it.
My naive intention was to use the cname as the publically
visible name, so I could move RT to another box later.

Well, sendmail itself translates the cname in To or From
into the arecord on the fly. So even though I properly
sent mail to rt@domainname, it arrived addressed to
rt@machinename.

I couldn’t find any way around this. So for admin reasons
here, I’m going to assign a separate IP (with arecord).
The box will have it’s machinename identity with one IP+arecord
and a second identity as rt with a separate IP (that I can
then move to another box.)

bobg

Having a problem with a few non local LAN individuals replying to
tickets. For some reason the reply to address for some users is
rt@machinename instead of rt@domainname.

What reply-to address? Looking at the actual emails RT sends to these
not-local-LAN people, what are their headers? Is the From address
correct? Is there a Reply-To? A Sender?

seph

Don’t know about postfix, but I ran into something similar
recently. I configured a machine whose dns arecord was
“machinename” and a cname of “domainname” pointing to it.
My naive intention was to use the cname as the publically
visible name, so I could move RT to another box later.

Well, sendmail itself translates the cname in To or From
into the arecord on the fly. So even though I properly
sent mail to rt@domainname, it arrived addressed to
rt@machinename.

I couldn’t find any way around this. So for admin reasons
here, I’m going to assign a separate IP (with arecord).
The box will have it’s machinename identity with one IP+arecord
and a second identity as rt with a separate IP (that I can
then move to another box.)

Yes, this is normal sendmail behaviour (sort of - there are a variety of
ways to control what it does with those addresses, depending heavily on
your environment).

There are too many different variables controlling what gets stuck
there to give a proper answer for what you should do to fix this, but
one observation is that sendmail expects DNS to be configured properly.
You should have an MX record for the machine, and an MX record
shouldn’t point to a CNAME. You can work around this by adding the CNAME
to the w class, something like

Cwfoo.uic.edu

where ‘foo.uic.edu’ is whatever domainname you want to use. But that’s
not ideal.

The Right Thing To Do™ is to fix your DNS and create an m4 config with
proper DEAFAULT_HOST, and add PSEUDONYM entries as needed.

Sendmail has a steep learning curve, and if you don’t want to climb it,
postfix is a great alternative with comparatively friendly
configuration. I personally like sendmail, but that might be Stockholm
Syndrome. And a careful look at headers will show that our company network
uses postfix now.

-j

Jamie Lawrence jal@jal.org
There are no answers, only cross-references.

rtusers-lists-fsck-com@jal.org wrote:

Sendmail has a steep learning curve, and if you don’t want to climb it,
postfix is a great alternative with comparatively friendly
configuration. I personally like sendmail, but that might be Stockholm
Syndrome. And a careful look at headers will show that our company network
uses postfix now.

While we’re using Postfix now too, I have to disagree about
configuration – I find a Sendmail m4 config at least as easy. An,
after all, even Eric Allman says cf files are “email assembly language”
not for 99% of us to use directly.

While we’re using Postfix now too, I have to disagree about
configuration – I find a Sendmail m4 config at least as easy. An,
after all, even Eric Allman says cf files are “email assembly language”
not for 99% of us to use directly.

I absolutely agree - the m4 config is just fine. It has been my
experience, though, that admins (myself included) don’t follow best
practice all the time, and modify things directly, leading to cascading
failures. While one can wag fingers, using a tool that doesn’t have
an intermediate step for configuration while still allowing direct
modification of the resulting config has a certain tendency to cause
failure in environments without strictly enforced policy.

In any case, I was only editorializing a bit, not trying to throw any
critical words.

-j, who learned all he needed to know about sendmail from the
towers-of-hanoi hack, and too much Bat book on an empty stomach.

S49
RHANOI:$+ $:1 2 3$1
R$-$-$-$[$+] $:$1$2$3$4
R$-$-$- $@$1$2$3
R$-$-$-@$
$:$>49 $1$3$2$4
R$-$-$-$* $:$>49 $2$3$1$4[Move Top Disk Of Peg $1 To Peg $3]
R$-$-$-$* $:$3$2$1@$4

Jamie Lawrence jal@jal.org
It is much easier to suggest solutions when you know nothing
about the problem.

After a archive search I did see the DNS issue and sendmail but do not
feel this applies, I have an A and MX record configured in dns for this
machine, it only happens for certain users or domains (not sure).

Here are some doctored headers

Managed-BY: RT 3.0.7_01 (Request Tracker... So much more than a help desk — Best Practical Solutions)
MIME-Version: 1.0
In-Reply-To: rt-379@vSupport
X-Mailer: Perl5 Mail::Internet v1.58
Sender: apache apache@machinename.domain.com
RT-Ticket: vSupport #379
Message-Id: rt-3.0.7_01-379-4715.6.22584084953232@domain.com
Precedence: bulk
Content-Type: text/plain; charset=“utf-8”
X-RT-Loop-Prevention: vSupport
X-RT-Original-Encoding: utf-8
RT-Originator: user@somecompany.com-----Original Message-----
From: seph [mailto:seph@directionless.org]
Sent: Tuesday, December 30, 2003 12:36 PM
To: Dan Fiorito
Cc: rt-users@lists.fsck.com
Subject: [rt-users] Re: Reply to address

Having a problem with a few non local LAN individuals replying to
tickets. For some reason the reply to address for some users is
rt@machinename instead of rt@domainname.

What reply-to address? Looking at the actual emails RT sends to these
not-local-LAN people, what are their headers? Is the From address
correct? Is there a Reply-To? A Sender?

seph
rt-users mailing list
rt-users@lists.fsck.com
http://lists.bestpractical.com/mailman/listinfo/rt-users

Have you read the FAQ? The RT FAQ Manager lives at http://fsck.com/rtfm

Here are some doctored headers

Sender: apache apache@machinename.domain.com
Reply-To: queue@machinename.domain.com
From: “Queuename via vSupport” queue@machinename.domain.com

My first guess is that some clients ignore the Reply-To, and reply to
the Sender.

seph

I can not find anything common yet, one user sits in an Exchange 5.5
environment, another in an Exchange 2000 environment. A full assortment
of Outlook clients sit in both places as well and all do the same thing.
I have other users in the same setups that work perfectly.

The clients reply to address is correct except that the machine name is
added and it is only in these 2 locations. I am stumped at this point.-----Original Message-----
From: seph [mailto:seph@directionless.org]
Sent: Tuesday, December 30, 2003 5:50 PM
To: Dan Fiorito
Cc: rt-users@lists.fsck.com
Subject: Re: Reply to address

Here are some doctored headers

Sender: apache apache@machinename.domain.com
Reply-To: queue@machinename.domain.com
From: “Queuename via vSupport” queue@machinename.domain.com

My first guess is that some clients ignore the Reply-To, and reply to
the Sender.

seph

I can not find anything common yet, one user sits in an Exchange 5.5
environment, another in an Exchange 2000 environment. A full assortment
of Outlook clients sit in both places as well and all do the same thing.
I have other users in the same setups that work perfectly.

I’d get rid of the Sender header and see if the problem goes away. Or
send a bunch of carefully constructed test messages to the clients to
observe their behavior.

seph

seph wrote:

Here are some doctored headers

Sender: apache apache@machinename.domain.com
Reply-To: queue@machinename.domain.com
From: “Queuename via vSupport” queue@machinename.domain.com

My first guess is that some clients ignore the Reply-To, and reply to
the Sender.

Specifically, Outlook turns:From: x@y.com
Reply-To: z@y.com
Sender: abc@foo.org

…into:

From: abc@foo.org (on behalf of x@y.com)

…and sends replies to abc@foo.org.