claw@kanga.nu said:
I potentially have the same problem -- the system occassionally gets very high load spikes, NOT due to MailMan, and Exim is configured to refuse connections when load gets too high.
A solution to this is to add your mailman host (which would typically be 127.0.0.1) into smtp_reserve_hosts and then set the related parameters such as smtp_accept_reserve, smtp_load_reserve and the various *_load* options so that incoming SMTP is always accepted from mailman (although the messages may be queued rather than immediately delivered).
The upside of this is that it puts less load on the system than additional invocations of exim from the command line, and all your load policy is pretty much in one place. Whichever method you use there may still be errors caused due to (for example) running out of spool space.
[I'll try and work out a reasonable note on this for the mailman/exim howto information - alternatively I would just *love* it if someone would come up with some text for me :-) ] http://www.exim.org/howto/mailman.html
Nigel.-- [ - Opinions expressed are personal and may not be shared by VData - ] [ Nigel Metheringham Nigel.Metheringham@VData.co.uk ] [ Phone: +44 1423 850000 Fax +44 1423 858866 ]
"NM" == Nigel Metheringham <Nigel.Metheringham@vdata.co.uk> writes:
NM> A solution to this is to add your mailman host (which would
NM> typically be 127.0.0.1) into smtp_reserve_hosts and then set
NM> the related parameters such as smtp_accept_reserve,
NM> smtp_load_reserve and the various *_load* options so that
NM> incoming SMTP is always accepted from mailman (although the
NM> messages may be queued rather than immediately delivered).That would be fine. Anybody know of similar parameters for Postfix or other MTAs? It makes sense that MTAs should be able to treat localhost connections differently than remote connections.
NM> The upside of this is that it puts less load on the system
NM> than additional invocations of exim from the command line, and
NM> all your load policy is pretty much in one place. Whichever
NM> method you use there may still be errors caused due to (for
NM> example) running out of spool space.And Mailman could have a bug, or Python could throw MemoryErrors, etc.
Anyway, if Mailman can't successfully do the SMTP transaction, and it gets exceptions it doesn't know how to deal with, the exception should at least get logged. I know that doesn't make you feel great, and the situation should be improved.
-Barry
participants (2)
-
bwarsaw＠cnri.reston.va.us -
Nigel Metheringham