Bounce detection

Hi,

I am new to rt and trying to get my head around everything (to the guys
who are developing it - TOP JOB!)

Something I notice is missing is any bounce detection, I also noted from
readiong the list archive alot of people have written their own little
customizations - so I was hoping someone had done it already, or can give
me some pointers.

I will have a look at doing this myself, in case that is the path I will
have to go down I will outline what I though would suffice (at least for
now).

Set envelope sender to something like: bounce+#@domain.com (# is the
ticket number email is being sent about). Then extend mail gate to have a
"bounce" service…which would very much be a comment, just how it
detects ticket number would be different. Then insert the message as a
comment.

This got me to thinking about permissions, and perhaps setting up a “mail
user” or something similar that has permissions to comment on tickets.

I know this is not the complete bounce detetion people would probably be
thinking of, but at least this would notify the ticket that a message
bounced - even if the users of that ticket need to manually figureout who
bounced and why…on a site that might have a good number of queue’s it
is better than the postmaster having to do it.

Comments, thoughts - even better has someone done this(or something
similar)? Even if is a pointer on say modifying the envelope sender any
advice would be appreciated. For the record I am running 2.0.15.

Cheers,

Stewart

Following this up, I seem to have managed to put some very loose bounce
detection in place.

Not being familiar with RT, I am not so sure what I have done will work.
Perhaps someone here might be able to shoot me out of the water and tell
me this will not work on a production box.

In short the files I modified were Interface/Email.pm Action/SendEmail.pm
and rt-mailgate.

I have not tested these AT ALL apart from seeing that things worked. But
here is what I saw happening on my test system.

Interface/Email.pm seemed to be used when an error occured and we needed
to say go away. so I set a -f postmaster@mydomain.com so that all messages
from errors could bounce to me.

Action/SendEmail.pm seemed to be called for generation of other messages
so this is hacked to set a return path of bounce+ticket-#@mydomain.com

rt-mailgate had a few lines added to figure out which ticket a bounce was
from.

finally I added a line to my sendmail aliases:
bounce+*: “|/usr/lib/request-tracker/rt-mailgate --queue sysgroup --action
comment”

The few tests I performed seemed to have worked.

I am hoping that others see my abouve hacks as viable in a production box,
but more importantly, if there is a reason for the above to not work - I
have only spent 2 hours with the RT code so my assumptions on how RT works
may very well be up the duff!

Cheers,

StewartOn Thu, 13 Mar 2003, Stewart James wrote:

Date: Thu, 13 Mar 2003 09:01:01 +1100 (EST)
From: Stewart James stewart.james@vu.edu.au
To: rt-users@lists.fsck.com
Subject: [rt-users] Bounce detection

Hi,

I am new to rt and trying to get my head around everything (to the guys
who are developing it - TOP JOB!)

Something I notice is missing is any bounce detection, I also noted from
readiong the list archive alot of people have written their own little
customizations - so I was hoping someone had done it already, or can give
me some pointers.

I will have a look at doing this myself, in case that is the path I will
have to go down I will outline what I though would suffice (at least for
now).

Set envelope sender to something like: bounce+#@domain.com (# is the
ticket number email is being sent about). Then extend mail gate to have a
“bounce” service…which would very much be a comment, just how it
detects ticket number would be different. Then insert the message as a
comment.

This got me to thinking about permissions, and perhaps setting up a “mail
user” or something similar that has permissions to comment on tickets.

I know this is not the complete bounce detetion people would probably be
thinking of, but at least this would notify the ticket that a message
bounced - even if the users of that ticket need to manually figureout who
bounced and why…on a site that might have a good number of queue’s it
is better than the postmaster having to do it.

Comments, thoughts - even better has someone done this(or something
similar)? Even if is a pointer on say modifying the envelope sender any
advice would be appreciated. For the record I am running 2.0.15.

Cheers,

Stewart


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

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

Email.pm.diff (655 Bytes)

SendEmail.pm.diff (816 Bytes)

rt-mailgate.diff (1.4 KB)

Woops,

Left a print(‘JJJ’) in the rt-mailgate diff.

either delete it or create a user ‘bounce’ and replce it with:
$CurrentUser->LoadByName(‘bounce’);

Make sure user bounce has permission to comment on tickets globally.

Stewart

In short the files I modified were Interface/Email.pm Action/SendEmail.pm
and rt-mailgate.

Interface/Email.pm seemed to be used when an error occured and we needed
to say go away. so I set a -f postmaster@mydomain.com so that all messages
from errors could bounce to me.

You don’t want to hard-code addresses inside the library files. Please
consider using $RT::OwnerEmail (set in the RT2 configuration file
etc/config.pm, and can be added to the RT3 config files for the same
effect)

Action/SendEmail.pm seemed to be called for generation of other messages
so this is hacked to set a return path of bounce+ticket-#@mydomain.com

Again, try to avoid hard-coding site-specific information into your
library files, and grab them from the configuration file.

rt-mailgate had a few lines added to figure out which ticket a bounce was
from.

finally I added a line to my sendmail aliases:
bounce+*: “|/usr/lib/request-tracker/rt-mailgate --queue sysgroup --action
comment”

There was a similar hack to do just this some time ago, which included the
relevant sendmail magic to supply the appropriate recipient address in the
command line option ( its more reliable if you are passed the address
through a --flag, rather than trying to figure it out by looking for a
‘To’ line in the message, which may not match ).

I am hoping that others see my abouve hacks as viable in a production box,
but more importantly, if there is a reason for the above to not work - I
have only spent 2 hours with the RT code so my assumptions on how RT works
may very well be up the duff!

‘By George, I think you’ve got it’ :wink:

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

                         Bruce Campbell                            RIPE
               Systems/Network Engineer                             NCC
             www.ripe.net - PGP562C8B1B             Operations/Security