Question on reducing load on MailScanner machine

[Julian Field]

On 25/06/2009 20:28, Christopher Fisk wrote:
>>   The previous thread about this didn't go very far because
>>   I suspect
>>   no-one is brave enough to actually try this.  Most of us
>>   just either
>>   optimise our installations to prevent the queue build-up
>>   in the first
>>   place or just add another box - it's the far less
>>   dangerous and the most
>>   travelled path.  That's also the reason it's not covered
>>   in the book.
>>      
> I can add another mx, but would prefer to have a single entry point for email on the network.  I think I would upgrade to a single more powerful machine rather than adding additional boxes for incoming messages.  We're running a relatively slow single processor machine at the moment and have a backup server that has been powered off sitting under it.
>    
The setup used by everyone else in the universe is to have multiple MX 
servers sharing the incoming mail load.
>    
>>   E-mail isn't instant messaging; a queue of 300 would
>>   impose nothing more
>>   than a few minutes delay at most which is perfectly
>>   acceptable to most
>>   people here.
>>      
> I did a typo, the queue is reaching 3000+.  Sorry about that.  Your point still remains, but you would be suprised at how many of our customers send an email to someone while on the phone and having even a few minute wait annoys them.  I'd rather throw hardware at the problem to get the queue down to 0.
>    
Let me get this straight.
You've got a huge mail queue, and yet you have a server sitting there 
switched off.
Dare I suggest you switch it on?
>
>    
>>   You're welcome to try NFS mounting your 'hold' directory
>>   and running
>>   another box on it at the same time; but you get to keep
>>   all the pieces
>>   if it breaks and to answer the phone to your users when
>>   it goes wrong
>>   and they get duplicate messages delivered to them or if
>>   their important
>>   mail get nuked.
>>      
> There is actually a good chance I will test this out.  If I do I will inform of the results.
>    
It won't work. Anyone sane runs multiple MX servers :-)
>
>    
>>   So my recommendation would be to avoid this; but if you
>>   are going to try
>>   it - do it on virtual machines and test it thoroughly
>>   (and document it
>>   for others too if it works!).
>>      
> How do the various child processes of MailScanner know when another child process is scanning a message in the queue?  Does the parent process keep track?  It seems like this would logically work.
>    
It does it all via file locking. And that locking is the same method 
used by your MTA. And that is not designed to work across NFS. So don't 
waste your time trying :)

Jules

-- 
Julian Field MEng CITP CEng
www.MailScanner.info
Buy the MailScanner book at www.MailScanner.info/store
Follow me at twitter.com/JulesFM

MailScanner customisation, or any advanced system administration help?
Contact me at Jules at Jules.FM

PGP footprint: EE81 D763 3DB0 0BFD E1DC 7222 11F6 5947 1415 B654
PGP public key: http://www.jules.fm/julesfm.asc


-- 
This message has been scanned for viruses and
dangerous content by MailScanner, and is
believed to be clean.