Index: [Article Count Order] [Thread]

Date:  Thu, 10 Aug 2006 11:52:03 +0800 (SGT)
From:  patricko (at mark) staff.singnet.com.sg
Subject:  [coba-e:06303] Re: break-smrsh-loop.pl question - RFC
To:  "Brian N. Smith" <brian (at mark) nuonce.net>
Cc:  coba-e (at mark) bluequartz.org
Message-Id:  <Pine.LNX.4.44.0608101114320.19628-100000 (at mark) staff.singnet.com.sg>
In-Reply-To:  <018501c6bc26$471f71a0$cc64a8c0@solunetdb0612c>
X-Mail-Count: 06303

Hi,
  
  Ok. 
  Some background information. (I know that I quite a nag)
  Anyway...

  Local mail loop in BQ setup.


>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>


  Condition 1:
               Original /usr/local/sbin/vacation.pl
               answered all mails, even the Sender is a Bounced.
               ie. From: MAILER-DAEMON (at mark) somedomain.com

  Imagine mail notification email addresses, 
  ping - pong among themselves.

  Solution:
           Upgraded vacation.pl to ignore mail notification email 
           addresses.


>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>

  
  Condition 2:
              Composite attributes that caused yet another local mail 
              looping.  Mainly, It is due to weak design of sendmail
              in local mails progressing.
              

              1) /var/spool/clientmqueue/
                 [Proven] no timeout in clientmqueue - local queue.
                          What-if there are mails stuck in clientmqueue?


              2) and smrsh - Sendmail Restricted (Privilege) Shell
                 [Proven] the wrapper smrsh for vacation.pl break down
                          before the actual calling of vacation.pl

   Solution:
            My opinion, run "break-smrsh-loop.pl" in BQ swatch as Active 
            Monitor (AM). Currently, a quick and dirty way is to call
            the script at crontab.

    
     
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
[What I have done]
 

 Factor out, modification to existing setup.
 If not possible, minimum changes. 
 
 Therefore when I designed the "break-smrsh-loop.pl",
 I make an attempt not to change sendmail conf, regardless
 of how lousy sendmail design can be. 
 
 The default behaviour after smrsh call, is to call vacation.pl
 and the local loop begin when .db cannot be write.

 "break-smrsh-loop.pl" will break the loop by making sendmail return
 file: " .<username>.db" not found and unable to lock 
 ERROR instead of the default ERROR (which is .<username>.db found, 
 able to lock but unable to write)
  



~ Request for any Comments. - RFC



Cheers
patrick





On Wed, 9 Aug 2006, Brian N. Smith wrote:

> Patrick,
> 
> I was looking at adding your "break-smrsh-loop.pl" into the BQ Core, but I am not sure how / where it gets called.  Obviously a symlink from /etc/smrsh, but how does sendmail call it?  Do you have the lines from the sendmail.mc you can send?
> 
> Please let me know!
> 
> --
> Brian N. Smith
> NuOnce Networks, Inc.
> =-=-=-=-=-=-=-=-=-=-=
> http://www.nuonce.net
> P: (321) 574-5227
> F: (321) 574-0554
> E: brian (at mark) nuonce.net