On Tue, 25 May 1999, Per Steinar Iversen wrote:
> mailstats is much more terse, it is also very quick as the statistics
> are accumulated by sendmail as it runs.
ph10@??? said:
> Exim can't do this, because it does not have a single, central,
> co-ordinating process. That is part of the reason why it scales well.
> It could, of course, update some file or other, but then there might
> be a bottleneck between the Exim processes queueing for access to the
> file.
Another alternative, which I started to implement a while back but did
not complete (unfortunately the source for that is with and owned by
Planet) is to sit a process on the log file and collect stats on the
fly.
This has a number of advantages:-
- these stats can be put into an MRTG/RRDtool RRD database
allowing them to be graphed easily
- it is easy to modify the process to collect the stats you care
about without risking breaking exim
- changing the backend is easy - ie SNMP queryable stats.
also just updating a file every n minutes with current
data.
Also in perl this is *very* easy to do - the File::Tail module (from
memory - I may have the name wrong) makes it very easy to camp on the
end of the log file - even handling log rollover transparently.
> It is now in beta test. There seems to be a command line and a Perl
> interface, but no C library (as far as I can see at a quick glance).
> As there seems to be some interest in maintaining Exim statistics on
> the fly, I have added this suggestion to the Wish List.
Doing this within exim with RRD could be a problem - I am not sure its
up to "increment the data value for this time" operations.
Nigel.
--
[ Nigel Metheringham Nigel.Metheringham@??? ]
[ Phone: +44 1423 850000 Fax +44 1423 858866 ]
--
*** Exim information can be found at
http://www.exim.org/ ***
