Hi list,

I'm running a cute little Postfix/Amavisd-new/whatever-cluster. For
some reason (better scalability, easier management, maybe geographic
distribution...) I'm right now preparing a lot of changes for our
platform.

As some of my ideas could be useful for others and as I would really
appreciate your feedback / input on some points I decided to share
my wired thoughts and doubts on this list.

To avoid an endless text monster I'm splitting my post in more parts:

   (1): Platform description
   (2): Database partitioning
   (3): Log aggregation


(1) Platform description

Ok, so let's start with a quick introduction to my (current) platform.
The front MX servers (right now there are two of them) are hit by 2-5
million delivery attempts a day, sometimes even more. The whole mail
filter cluster is build out of not more than 3-4 hosts running something
like 30 vServers these days. This setup should easily be able to handle
the current load many times.

We prefer to dedicate one virtual server to each service - and to run
at least two instances of each service (even if a single one could
easily handle the load). Doing things this way is more work at the
beginning, but it makes it easier to scale out once required. And it
allows me to shut down each machine immediately without trouble if for
whataver strange reason I have the need to do so.

The "cluster" consists of DNS caches, database servers, amavis servers,
front MX servers, some queue servers behind the filters with different
queuing behaviour, responsible to statically route mails to our own
mailserver(s) or to those of our customers, log servers. As an ISP we
have to also take care of dinosauriers like ETRN etc.

Amavis and Postfix servers are right now using the same MySQL "cluster",
consisting of one master and two slaves - handling a monthly medium of
3-400 queries/second on each of them.

Issues I would like to address in the ongoing migration step:

- Garbage collection
- Geographic distribution / disaster recovery
- Scalability (even if my current setup will never be THAT large ;-)

So, that's it (more or less). I feel sorry for my broken English, I'm
trying to give my best! You'll find technical details and questions in
the following two mails.

Regards,
Thomas Gelf


-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone
_______________________________________________
AMaViS-user mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/amavis-user
AMaViS-FAQ:http://www.amavis.org/amavis-faq.php3
AMaViS-HowTos:http://www.amavis.org/howto/

Reply via email to