[...]
Lastly, could someone recommend some relatively cheap third party tracking system (pixel tracking?) that we can use for a month of two to verify if your logging system is not working properly?
If you find a discrepancy between a 3rd party system and your own log files, the one thing that you can know with 100% certainty is that the problem is with the 3rd party system. Don't waste your time.
While this is of course true, you can not always trust the content of your logfiles to reflect reality. E.g. apache have some known bugs[1] that will render your statistics way to high and thus useless in some cases and I've seem similar behavior with proftpd. It puts _way_ to high (exabyte) values in the "bytes sent" field of the log, but I have not found it to be a confirmed bug.
So if your statistics differ from your log files, then it is obviously the analyzers fault.
If it differ from what makes sense however, it may be (has been in my case) the fault of the program generating the logs.
The Apache bug you reference is what you might reasonably call an exception - if aborted transfers make up a significant proportion of
Unfortunately it does. Just for a test, I tried to exclude lines with partial requests (code 206), and the two reports differed enourmously. IIRC the one with part.reqs. included was about 30 times as big as the other, which was more in line with what was excpected.
your logfile entries, then you probably have bigger problems than a discrepancy in your logs. (If you're being charged for bandwidth, and
I blame it on badly written "download accelerators" rather that transfers that got "incorrectly" aborted.
you have a very large file available for download that is cancelled a lot, then you might care). But there's no way that any 3rd party "web bug" can give you anything at all about the bandwidth used by your site, so it's not much of a counter argument.
Web bugs are always going to give you an approximation at best - try sticking 3 web bugs from 3 different "web statistcis tracking" services on your website, and you'll see just how "approximate" they are!
You're probably right (haven't tried), so for a true total amount of used bandwidth one should probably measure the bytes passing the switch/router. (We only do this for bps values, but I guess it can't be that hard to have mrtg keep track of the total sum too.)
Best regards, Emil
+------------------------------------------------------------------------ | TO UNSUBSCRIBE from this list: | http://lists.meer.net/mailman/listinfo/analog-help | | Usenet version: news://news.gmane.org/gmane.comp.web.analog.general | List archives: http://www.analog.cx/docs/mailing.html#listarchives +------------------------------------------------------------------------