[logs] Error messages from syslogd
Russell Fulton
r.fulton at auckland.ac.nz
Wed Jul 11 14:55:14 PDT 2007
Marcus J. Ranum wrote:
> Justin Mitchell wrote:
>
>> Regular expressions, experience, and research are priceless (w/ an emphasis on
>> regular expressions).
>>
>
> #include <rants/mjr-why-regexps-are-the-wrong-thing.h>
>
>
for those who missed the original
http://lists.jammed.com/loganalysis/2002/06/0015.html
Marcus: I've reread that post and largely agree with it (although,
being a rant some points are a bit over stated ;). I am working on a
log management system written in ruby and yes, it uses REs. I'll add
one other bitch to your list -- SELMS (Simple Extensible Log Management
System) keeps all the patterns in a structured configuration file and RE
are a pain to parse. The way I got around it was to terminate REs with
a <tab> and grab every thing from the start of the RE to the <tab> and
pass the lot to Ruby to check (and return the compiled version as an
object which is then stored or an error message). It works but isn't
beautiful. I digress...
Given a different expression matching library it would be reasonably
straight forward to add it to SELMS -- that is what the Extensible bit
is about ;)
So does anyone have such a matching library that they can Open Source?
Even some thing that I can start from? Libpcre if all else fails I
guess? Hmmm... may be I'll talk to my friends in CS and Software
Engineering -- that might make a good student project.
I will be releasing SELMS later this year -- it is an out growth from my
sl2 and sl3 scripts that are already linked in the log analysis web
site. If anyone is using either of them then SELMS is pretty well a
drop in replacement but does much more. I have just added 'real time'
alerting, as opposed to the periodic 'artificial ignorance' stuff which
is similar to sl3. Once we have the real time stuff running in
production for a few months I'll put it up on RubyForge and announce it
in the obvious places.
Russell
More information about the LogAnalysis
mailing list