Text written by Nagendra Mishr at 12:48 PM 10/7/99 -0400:
>
>Does anyone know the best way to monitor running Systems? I.e. I want
>to know that somthing is wrong with the system... Is the best way to
>read the logs?
Depends on your situation. If you have multiple machines, and especially if
you have a shell account on some machine in a different location, you might
want to run something like Big Brother.
Big Brother is a set of shell scripts that run periodic tests on a variety
of services (including POP3 and SMTP), give a display of the situation on a
Web page, and can even page you if something goes wrong.
In particular, for its testing of services, it actually tries to form an
SMTP or POP3 connection to the machine being monitored. This means it's
doing real tests; if this script can't make an SMTP connection, then it's
likely that real users can't, either. I really like this as a monitoring
method. In general, where possible, I really like the idea of remote
testing on the actual ports used rather than simply looking at the logs; I
feel that you look at the logs *after* something goes wrong, so you can
figure out what it was. But some errors won't show up in the logs.
Additionally, Big Brother has an optional client portion that sits on the
monitored machine and lets you know if any of certain listed (configurable)
processes aren't running.
I can't give it a completely untarnished recommendation; I do have some
problems with Big Brother. For one, I think it should be reimplemented in
Perl; many of the text-manipulation and similar tasks it performs would be
so much cleaner there. And it's not as versatile or flexible as it could
be; trying to bend it to do things that are similar to, but not quite, what
it already does always seems harder than it should be. I would definitely
not recommend it for a large and complex monitoring situation, and I would
caution anyone evaluating it that while it's pretty nice as long as you
only want the functionality that already comes with it, it gets crotchety
and brittle when you alter it.
However, you might want to look at the way it does things and then roll
your own. You can find Big Brother at http://www.maclawran.ca.
-----------------------------------------------------------------
Kai MacTane
System Administrator
Online Partners.com, Inc.
-----------------------------------------------------------------
>From the Jargon File: (v4.0.0, 25 Jul 1996)
hungus /huhng'g*s/ /adj./
[perhaps related to slang `humongous'] Large, unwieldy, usually unman-
ageable. "TCP is a hungus piece of code."