On 2012-05-30 at 13:06 +1100, Костырев Александр Алексеевич wrote: > I'm trying to squeeze max perfomance from all of the components, so i > use exim without any limitations.
This situation is an example of why trying to go as fast as possible can lead to going slowly. > /var/log/exim/main.log > > 2012-05-29 17:31:28 1SZFxo-0004rG-3r == [email protected] > R=dspam_router T=dspam_spamcheck_socket defer (104): Connection reset > by peer: NULL: Connection reset by peer You're overloading dspam to the point where it can't handle the rate. Perhaps you can tune it to raise the size of the listen queue? Make sure that you set the various queue_only_* options to spool the data without trying to deliver immediately, and then make sure that you don't have the -q/-qq options set too high, so that Exim won't start too many queue runners at once. Then your messages flowing into dspam will be rate-limited by the number of queue-runners you have. Eg, to speed things *up*, you'd use split_spool_directory and run the queue-runners from cron instead of Exim, with a script which runs queue runners per spool sub-directory. To slow down, you make sure you're not invoking too many queue runners at once. To go faster with a slow component, you run a pool of the slow component and tell Exim to use multiple addresses, with "hosts_randomize" set to distribute the load "probably evenly" over the available backends. -Phil -- ## List details at https://lists.exim.org/mailman/listinfo/exim-users ## Exim details at http://www.exim.org/ ## Please use the Wiki with this list - http://wiki.exim.org/
