I have managed to get Nagios up and running, but as a central server checking remote clients. So far I have it working on a Solaris 8 server with nine Solaris 8 servers and about forty Sun Blade 1500 and 2500 workstations.
Eventually I would like for it to monitor about 30 Solaris servers, 40 Solaris workstations, and maybe (if I can get management approval) about 200 Windows 2003 hosts - they’re looking at paying exhorbitant money for Microsoft MOM or SCOM - but maybe I can sway them to the open source Nagios.
I’ve been reading through this and other sites looking for the best way to implement this type of setup. With let’s say 500 hosts to monitor, I can see that the configuration files will be ginormous and maybe the server will start to overlap its checks.I noticed that Jakkedup is running a distributed environment with the clients sending polling data to the Nagios server - would this be the best way of setting up this kind of monitoring service?
I’ve read through all the documentation for Nagios 2.6 and I have found it to be excellent - it’s gotten me this far, but I am not fully understanding the distributed monitoring concept.
I guess my real question is, how to best plan and setup a monitoring environment for 3-400 hosts?
Concept is pretty straight forward. The central server doesn’t make the checks. The remote hosts send the output of the checks that they make, to the central server.
Windows server’s have a few options, and one of them is nc_net. NC_net runs the checks on the windows box and sends the output to the central nagios server.
What I like to do on 'nix type boxes, is setup nagios on that box, and let that nagios install make the checks. These 'nix boxes then send the output to the central nagios server. It’s not a resource hog, and we have done just that on many of our production 'nix systems to monitor practically anything you can think of.
No matter what kind of setup you choose, active (all checks made by one bo:evil: or passive (distributed setup), you will have to install “something” on the remote hosts if it’s a windows box (unless all you are going to do is ping it. If it’s a 'nix box, then likewise, something is going to have to be setup on it, snmpd, cronjob’s, or nagios setup. I chose nagios on remote 'nix boxes, since it gives the most flexibility.
What exactly don’t you understand about the concept?
For a distributed setup, will I have to compile from source on each of the distributed servers (the Solaris 8 OS is the same), but they are missing most of the libraries - gd, jpeg, png, etc. Since they won’t be generating any html pages or images, I would assume they can be skipped, but I’m not sure.
Currently the nagios clients have nrpe and v1.9 of the plugins installed. The only library that I had to install on the clients was libiconv.
Also, is it possible to run a semi-distributed seutp alongside a fully active setup on the same box with the same Nagios install, or would have to seperate the installs?
Thanks again for your response… I’m loving Nagios right now, it seems so much more comprehensive than Big Brother, which I’ve been running for the past 2 years here.