> -----Ursprüngliche Nachricht----- > Von: [EMAIL PROTECTED] [mailto:amavis-user- > [EMAIL PROTECTED] Im Auftrag von Clifton Royston > Gesendet: Mittwoch, 2. April 2008 23:52 > An: Rocco Scappatura > Cc: [email protected] > Betreff: Re: [AMaViS-user] How to justify amavisd-new delay > > On Wed, Apr 02, 2008 at 03:15:03PM +0200, Rocco Scappatura wrote: > > > > > Thanks Gary, > > > > > > you're explanation is very brief and fully explanatory. > > > > > > I surely will execute some measurement of SA efficiency by > > > increasing log verbosity. > > > > > > In the mean time I find very embarassing some situations > > > where I have a congtestioned SMTP gateway server and I can't > > > avoid new incoming SMTP session from being managed by that > > > server (In this manner I only increase congestion. While I > > > like to let that server to 'digest' the queued mail traffic > > > before accepting new SMTP connection, and redirect the new > > > SMTP connection on other SMTP gateway server of my mail system). > > > > > > I use DNS balancing and when a queue on a server is > > > congestioned I get no effect from taking away that server > > > from being pointed by an MX record, because DNS changes > > > propagations take long time. :-) > > > > > > Is there any trick to tackle this issue? > > > > Any answer will be appreciate, even if it not came directly from > Gary.. > > Heh. > > As you have found, it's very difficult to do anything other than very > crude load management via DNS. Obviously you can make it a little bit > better by keeping your DNS TTLs short, but it's a very blunt tool. > > The only load-balancing method that provides a more immediate and > tunable response is to use a different (non-DNS) approach, such as a > hardware load-balancer (I <3 my Foundry ServerIron) or some > routing-based software system such as CARP. > > When one server is congested, your best bet and simplest soulution > may be to simply suspend the incoming SMTP by editing the postfix > master.cf file, temporarily comment out the line which has Postfix > listening on port 25, and reload Postfix. This is a drastic measure, > but I've used it from time to time to manually allow one server to > clear a big backlog. Once the backlog has cleared, you *MUST* remember > to uncomment the line and reload Postfix again, so that it can resume > accepting mail.
Since the bottleneck is SA and not Postfix you should rather use load balancing for the internal forwarding to amavisd-new and distribute the load of content scanning to different machines. Set in master.cf of postfix for example -o content_filter=amavisd-new:amavis.domain.tld:10024 amavis.domain.tld can also be a MX record. If all amavisd-new processes on the machine with the highest priority are in use, postfix should automatically forward the mail to the box with the successive priority. With $forward_method = 'smtp:*:10025'; in amavisd.conf amavis will forward the scanned message back to the originating postfix server and you can share several amavis installations on more than one SMTP gateway making it possible to use also round robin for your SMTP gateways. Greetings Bernd ------------------------------------------------------------------------- Check out the new SourceForge.net Marketplace. It's the best place to buy or sell services for just about anything Open Source. http://ad.doubleclick.net/clk;164216239;13503038;w?http://sf.net/marketplace _______________________________________________ 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/
