Noel Jones wrote: > On 9/15/2009 11:21 AM, Clayton Keller wrote: >> Clayton Keller wrote: >>> Noel Jones wrote: >>>> On 9/7/2009 1:55 PM, Mark Martinec wrote: >>>>> Clayton, >>>>> >>>>>>>>>> I'm running amavisd-new 2.3.3 and it now refuses connections. It is >>>>>>>>>> up >>>>>>>>>> and running according to ps but postfix cannot talk to it. The >>>>>>>>>> postgres-db it uses is also up and running. Any ideas where to start >>>>>>>>>> looking? >>>>>>>>> Start with the basics... what's in the log, can you connect to the >>>>>>>>> amavisd port with telnet, etc. >>>>>>>> It's reachable by telnet. The only things that get logged by amavis is >>>>>>>> when it is restarted. Postfix says connection is refused. >>>>> Postfix keeps a memory of next-hops which appear to be down, failing >>>>> several attempts to connect. This information is cleared after a while, >>>>> but until it does clear, postfix can claim in the log the mailer is >>>>> unreachable, even though it came up by now. >>>>> >>>>> This usually happens when amavisd is restarted while Postfix is busily >>>>> feeding mail to it. It seems the longer amavisd restart takes and the more >>>>> sessions are open, the higher the chances of Postfix treating the amavis >>>>> feeds as being down. >>>>> >>>>> Eventually postfix would notice the service is up again, but to rush >>>>> the recovery, a quick and ugly solution is to restart postfix, >>>>> so that it forgets the condition. A 'postfix flush' may help too. >>>>> >>>>> When activity is low and a restart doesn't take long, the above drastic >>>>> action is unnecessary. So it is best to avoid restarting amavisd during >>>>> busy hours. So far I haven't come across a better solution - the question >>>>> perhaps better belongs to the postfix-user mailing list. >>>>> >>>>> Mark >>>>> >>>> Running "postfix reload" on a busy postfix system can make >>>> things worse. All mail - both deferred and active - must be >>>> moved to the incoming queue, then moved back to the active >>>> queue for a fresh delivery attempt. This can cause a huge >>>> spike in disk activity bringing the system to a near-halt >>>> while the files are moved around. Also, any truly >>>> undeliverable deferred mail will be retried, adding to the >>>> load. This isn't much of an issue on a lightly loaded system, >>>> but becomes more of a problem the more mail waiting in the queue. >>>> >>>> Suggested settings for tuning high-volume destinations for >>>> fault tolerance can be found under >>>> http://www.postfix.org/QSHAPE_README.html#backlog >>>> >>>> Assuming postfix 2.5 or newer, and the amavisd-new master.cf >>>> transport is named "amavisd", setting should look something like: >>>> # main.cf >>>> amavisd_destination_concurrency_failed_cohort_limit = 100 >>>> amavisd_destination_concurrency_limit = 20 >>>> Note: the values suggested above are rather arbitrary, some >>>> sites may need to adjust them. See postfix docs for details. >>>> >>>> With postfix 2.4 and earlier, you can use: >>>> # main.cf >>>> amavisd_initial_destination_concurrency = 2000 >>>> amvaisd_destination_concurrency_limit = 2000 >>>> Note: the "2000" is a rather arbitrary number, large sites may >>>> need a higher value. See postfix docs for details. >>>> >>>> -- Noel Jones >>>> >>> Mark/Noel >>> >>> Thank you both for the added information. >>> >>> I had not seen the xxx_destination_concurrency_failed_cohort_limit and >>> xxx_destination_concurrency_limit related configuration options in the >>> documentation. This may be the avenue I need to approach this from. I >>> will see what I can find and continue my own troubleshooting on my issue. >>> >> Noel, >> >> I'm not seeing those options in the 2.3.x versions. Could I be >> overlooking them. >> > > Yes, this will work on postfix 2.3 using the "2.4 and earlier" > settings mentioned above. The settings won't show up in > "postconf" output. > > See postfix docs or there are several threads about this in > the postfix-users archive. > > -- Noel Jones >
Thanks. I apparently chose not to read over the "and earlier" my apologies. ------------------------------------------------------------------------------ Come build with us! The BlackBerry® Developer Conference in SF, CA is the only developer event you need to attend this year. Jumpstart your developing skills, take BlackBerry mobile applications to market and stay ahead of the curve. Join us from November 9-12, 2009. Register now! http://p.sf.net/sfu/devconf _______________________________________________ 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/
