[3079] in Release_7.7_team
Re: A better registration system.
daemon@ATHENA.MIT.EDU (Greg Hudson)
Thu Jan 3 18:38:43 2002
Date: Thu, 3 Jan 2002 18:38:38 -0500 (EST)
From: Greg Hudson <ghudson@MIT.EDU>
To: Bill Cattey <wdc@mit.edu>
cc: <release-team@mit.edu>
In-Reply-To: <wwBCWUhz0001BoQJcB@mit.edu>
Message-ID: <Pine.LNX.4.30L.0201031828480.16359-100000@egyptian-gods.mit.edu>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
On Thu, 3 Jan 2002, Bill Cattey wrote:
> I know that Athena Suns do a syslog every time they reboot.
All of our platforms, actually.
> I wonder if there could be some sort of syslog specifically for every
> Athena install, that would then eventually turn into a report of X
> systems on subnet Y over time period T.
> What do people think about the feasibility of this?
After an install, a machine might get moved, renamed, discarded, or
reinstalled with something else. The boot-time syslog we have is
more reliable, but not perfect; we get no false negatives due to machine
moves, but still get some false positives from stale hostnames unless we
come up with a way of detecting that.
Because of dual-boot machines, it can be hard to detect if a hostname is
stale (was this machine reinstalled with Windows, or is it just booted to
it?). Probably the most reliable report would be based on something like:
* Count all Athena machines which booted in the last month, without
checking.
* Count all Athena machines which booted in the last year, but only if
they answer athinfo queries (and aren't Pismere machines).
That's not perfect, but I don't think we can easily improve on it. It
will give false positives on machines which were renamed,
discarded, or reinstalled with non-Athena in the last month, and will give
false negatives on dual-boot machines which haven't booted Athena in the
last month as well as machines which have been up a long time and have
athinfo disabled. Those should all be small categories, though.