Guy Hulbert wrote:
However:
~/tmp/dlog-0.9.8b$ wc *.[lyh12]
4598 11903 117593 total
wc qmailmrtg7-4.2/*.*
1077 2769 28913 total
wc tinydns-rrd-0.50/tinydns-rrd
345 1058 10014 tinydns-rrd-0.50/tinydns-rrd
Obviously the line counts are not a completely fair comparison but it
seems that the tools are progressively less accessible (perl, C, lex
+yacc) and the code seems to become *more* complex (though each tool
covers more different logs).
That's not really fair as dlog consists of analyzers for many other tools. The
data-collecting tools can be used directly to feed mrtg for instance:
$ wc dlogtiny.*
197 498 4216 total
$ wc dlogqmail.*
195 488 3889 total
But the prize winner will still be the dodlog.pl script that creates the graphs:
$ wc dodlog.*
1804 5324 64631 total
(Approx. 1/3 is definitions for the RRDtool databases that could exist in a
config-file maybe)
- but again that script handles all the different types of log files, but it
does not have to be executed on the same hosts as the data-collecting tools.
Many largescale installations produce GB's of logfiles, and the small binary
file that actually searches through the logfile puts minimum strain on the
production hosts (but of course it needs to traverse through the data once).
Then you can have your way with the output (produce graphs, or pass the info to
munin) at another host.
Is it really necessary to have a different program for each log or is it just
easier to do it that way ?
Many of the tools produce output that is very similar (I think axfrdns
and tinydns is almost identical), so it's difficult to make one pattern
to rule them all.
-Skaarup
