[Mimedefang] Greylisting Code on Wiki

[Jeff Rife]

On 12 Dec 2007 at 19:03, Paul Murphy wrote:

> One issue which I've seen is on low-usage systems (such as my home server),
> the MySQL connection times out, meaning that the slaves crash when the next
> message comes in to them.  I tried lots of things to stop this happening,
> including cached connections and checking the connection status and
> reconnecting if it wasn't live, but basically I couldn't avoid the slave
> dying.  However, this is no real problem - the multiplexor starts a new one,
> the sender tries again, and all is well, since this only happens on very
> lightly loaded systems, and I have 2 slaves in total at quiet periods.

I never have more than 2 slaves in my implementation, but never have 
these timeouts.  I suspect it's because I scavenge the greylist every 5 
minutes using the filter_tick hook in MD.  I also do opportunistic 
generation of 25-bit hashcashes during filter_tick, so that I can have 
them handy when the next outgoing e-mail to that recipient is sent.  
Since SpamAssassin in the default config puts some nice negative scores 
on the larger hashcash headers, it's a pretty good use of otherwise 
idle time.

I also have two MXs that both connect to the same database, so that 
makes it even more likely I have some real idle time.


--
Jeff Rife |  
          | 
http://www.nabs.net/Cartoons/ShermansLagoon/OtherWhiteMeat.jpg