On Wed, 4 Apr 2007, maillist wrote:
I have seen a few people present, on this mail list, nicely detailed graphs,
that obviously were the result of some server output, but they focused on
email, mainly spam. I am interested in having the same. Does anyone have any
recommendations for a good package that can do this?
All I currently use is logwatch. It's nice for my needs to administer, but
the boss would like to see something that he can understand without having to
do so much thinking. Maybe he wants to replace me with a bar-graph.
IMO, more statistics == better. Your boss would probably agree.
I use a heavily modified version of Mailgraph to get not just the
stock mailgraph stuff, but a bunch of other data, including the
effectiveness of our different RBLs, messages greylisted, etc.
I also use a heavily modified version of sa-stats to figure out which
of our rules are most effective, which hit the most spam/ham, etc.
I've also written a custom log analyzer to get data from ClamAV on
which viruses we're seeing the most, and a great big log analysis tool
to generate tons and tons of email statistics. You can see sample
output here: http://www.nebrwesleyan.edu/people/stpierre/spam-stats.html
(Note the downtime that MX node experienced last week.) The code is
pretty unpolished, and would really only be useful to someone with the
same setup as us, but it gives you an idea of things you might look at
graphing. The Perl is really pretty simple -- File::Tail,
Parse::Syslog, and GD::Graph are your friend in this endeavor.
I'd also recommend, if you end up writing your own tool, generating
hard numbers as well as pretty graphs. You can put the graph showing
the increase in spam|mail volume|whatever in your slideshow and
mention the hard numbers in your presentation on why you need N more
servers and X more sysadmins.
Good luck!
Chris St. Pierre
Unix Systems Administrator
Nebraska Wesleyan University
----------------------------
Never send mail to [EMAIL PROTECTED]
