nanog mailing list archives

Re: Monitoring Tools


From: Phil Regnauld <regnauld () nsrc org>
Date: Thu, 19 Aug 2010 22:23:44 +0200

Nathan Eisenberg (nathan) writes:
It hasn't really changed. Almost every monitoring package I've found
where you want to monitor something like 'disk space free on /' requires
a daemon of some sort on the host - whether that's SNMPD or their agent.

        Anything else than SNMP is a hassle (IMHO).

        I understand the idea of having a dedicated agent for some hosts -
        Windows for instance, when querying the WMI - and often it's the only
        way for a vendor to have a predictable, verified  element in the greater
        scheme (the network).

        But in most cases, monitoring can be achieved by extending the SNMP mib,
        and using and custom scripts that will report on mail queue size, in-house
        application status, etc...

FWIW, I have had their agent running on many, many servers over the
years - it has never caused me a moment of heartache (for safety's sake,
iptables restricts who can talk to the agent, which has its own control
mechanism built in to define who it will talk to, and it runs as a
restricted user, just in case).

        <hat employer=other>
        While developing our own monitoring product, we've had to deal with
        various constraints from the customer side, for instance pharmaceutical
        companies where there was no way installing an agent on PLC machines would
        pass internal audit, without having the entire system re-validated (we're
        talking FDA-validated medication production here).
        </hat>

        But often, SNMPD ships with or is available as an optional base
        component (Windows, most UNIXes) and it's easier to convince the IT
        suits.  Go figure.

        Oh, and it avoided us having to install an agent on 1000+ servers :)

        Cheers,
        Phil


Current thread: