Good Agentless monitoring packages?

[Roy S. Rapoport]

On Tue, Nov 25, 2003 at 05:12:32PM -0800, Michael T. Halligan wrote:
> Well basically I just want the ability to keep a customer's systems pretty
> vanilla, without adding much (or anything) to them that doesn't come with
> the default distribution (redhat 7.3 is their standard).  It becomes a lot
> easier to maintain consitency over the 200+ servers we'll be monitoring
> by not having to worry about installing newer versions of nrpe/nagios
> agents when we do upgrades, etc.. And NRPE agents get ugly when you
> begin monitoring servers behind several levels of firewalls, especially
> when a lot of these servers don't have gateways or routes out of their
> subnets (at that point an snmp collector or a script to periodically collect
> rpc stats become a lot simpler).
> 
> We definately want more features than mrtg has.. We're considering just nagios
> and biting the bullet on the agents, big sister, or cricket.. 

I played with Big Sister and found it ... well, not my cup of tea I
suppose is the best way I can put it.  I'd give you a technical critique of
it if I had any, but I just found it not comfortable to work with.  

I've deployed MRTG in an environment that was going to finish an HP
OpenView installation 'any day now' (and were thinking that way for
approximately two years) and then customized it using net-snmp's
extensibility capabilities to do some pretty nice things, but again that's
mostly on the trending, rather than the alerting side.  For alerting, I
find that Nagios' nuanced and granular approach to configuration and
alerting concepts is very hard to beat.

I also happen to hate deploying customized software on edge systems,
prefering to rely on (or deploy) more intelligence on the core systems, so
I completely agree with your preference (which, I'm sure, causes you great
relief :) ).  

What would you like to monitor? Have you tried building Nagios plugins?
They're trivially easy...

-roy