[473] in athena10
Re: counterlog design discussion
daemon@ATHENA.MIT.EDU (Timothy G Abbott)
Wed Aug 27 15:22:50 2008
Date: Wed, 27 Aug 2008 15:22:05 -0400 (EDT)
From: Timothy G Abbott <tabbott@MIT.EDU>
To: ghudson@mit.edu
cc: athena10@mit.edu
In-Reply-To: <200808271637.m7RGbfZK017153@outgoing.mit.edu>
Message-ID: <alpine.DEB.1.10.0808271452570.26686@vinegar-pot.mit.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; format=flowed; charset=US-ASCII
On Wed, 27 Aug 2008, ghudson@MIT.EDU wrote:
> In Athena 9.4 we have a little script called counterlog which phones
> home (via syslog) with the machine type, version, and a machine
> identifier which survives reinstalls and IP address changes. We use
> it to get counts of the number of distinct Athena machines which have
> been active over the last month.
>
> I'd like to implement something similar for Athena 10. The counts
> SIPB gets from debathena.mit.edu are decent but I believe they're
> distorted by machines using DHCP on different networks.
Just to be clear on how the debathena.mit.edu counts work, they actually
map IP addresses to MAC addresses count distinct MAC addresses. This will
double-count some machines that use more than one MAC address (i.e. use
both wireless and wired), but it's much more accurate than counting IP
addresses and in not particularly subject to distortion due to DHCP on
different networks (as can be seen by the comparatively huge IP address
numbers available for comparison on http://debathena.mit.edu/users).
> * Use wget or curl to post the data to a URL on the apt server
> (athena10.mit.edu or debathena.mit.edu as found in
> sources.list), and have a CGI script record it. Adds a little
> complexity to the web server setup but shouldn't be too bad.
I think I prefer this style of mechanism to writing a script to talk the
syslog protocol -- it saves us from having to respond to any changes that
might occur is said protocol.
It may be interesting to look at how the debian popularity-contest package
works.
-Tim Abbott
