Ralf Hildebrandt:
> * Victor Duchovni <victor.ducho...@morganstanley.com>:
> > On Thu, Feb 10, 2011 at 10:50:20PM +0100, Jeroen Geilman wrote:
> >
> > >> and I'm not sure how
> > >> smtp_connection_reuse_time_limit = 300s
> > >>
> > >> could be lowered in such a way that busy destination MXes are not
> > >> keeping a lot of mail in the active queue...
> >
> > The re-use time should equal or exceed the duration of the slowest
> > reasonably plausible connection + transaction latency when one of the
> > MX hosts is unwell (overloaded). If you drop all the other SMTP timers
> > you can drop the re-use time-limit to match,
>
> OK, then I should definitely drop that to a much lower value!!
Please, don't blindly scale down all time parameters.
The short timeouts for connect/helo are part of a DISCOVERY mechanism.
This helps you to discover which server is good or bad. But you don't
want to discover a bad server too often.
The re-use time-limit is part of a CACHING mechanism. This helps
you to remember which server is good, so you don't have to waste
time skipping over the bad ones.
Setting the re-use time-limit too short makes the caching mechanism
useless. Postfix is smart enough to discover that a cached server
goes bad, as long as you set smtp_rset_timeout properly.
Wietse
