Nagios question

[Red Squirrel]

I've been playing with Pandora and there are a few aspects of it I don't like. One is the way the threshold setting is very confusing, and also the fact that I cannot set the threshold to be high AND low, only one or the other. Ex: I can monitor a value and if it goes above 50 I get a warning, if it goes above 60 it goes critical. But I cannot monitor a value and make it warn if it's below 10 OR above 50 and make it go critical if it's below 0 OR above 60. Can do low, or high, but not both. Basically I rather set a range, and tell it to alarm if it falls outside of that range. Much simpler.

Also how easy is it to deploy agents? Is it a simple binary that will "just work" or some complex thing with tons of dependencies? That's the nice thing with pandora is the agent is very simple, and it's very easy to add custom monitor points like temperature from a text file as a single monitor point value is gathered from a command, so I can use sed/grep etc to grab a value from the Linux environment whether it's a text file or a command's output. Can nagios do this as well?

Also does Nagios use Flash? Pandora uses it, and flash locks up all the time on my system so that alone is reason enough to switch to something else.

 

[dave99]

no, basic nagios is plain html, no flash. You can probably make nagios do what you want by modifying a perl based check that someone has already built. There are some temperature ones out there that check high & low values. There are a variety of agents, nrpe, nsclient++, nc_net (I use this one), and probably more. Nagios can probably monitor just about anything you want, but the trick is you might have to write the check yourself in perl, python, php etc, anything you can get to return a result code of 0 to 3.

 

[Red Squirrel]

no, basic nagios is plain html, no flash. You can probably make nagios do what you want by modifying a perl based check that someone has already built. There are some temperature ones out there that check high & low values. There are a variety of agents, nrpe, nsclient++, nc_net (I use this one), and probably more. Nagios can probably monitor just about anything you want, but the trick is you might have to write the check yourself in perl, python, php etc, anything you can get to return a result code of 0 to 3.

Hmm so it does not actually see the real value? Does it have any graphing? That's one thing Pandora has that is kind of nice. It sees the value, and from there I decide what to do with it. So with Nagios I'm sending 0 to 3 only? (which I'm guessing is ok, minor, major, critical?)

 

[dave99]

nagios itself doesn't graph, but you can have check results come out of nagios into a graphing app like pnp4nagios, rrdtool directly etc. Nagios doesn't care/process what your check value is, it only cares about the return code, but you can have eventhandlers that fire based on a warning or critical result. For example, I have an app on a linux box that tends to fail on occasion, the server check has an eventhandler that will ssh to that box and restart the service when it fails rather than annoy me with it.

I integrate nagios and cacti together (cacti for graphing and trending), it works pretty well.

The result codes are:
0 = ok
1 = warning
2 = critical
3 = unknown/unreachable

 

[obrith]

nagios itself doesn't graph, but you can have check results come out of nagios into a graphing app like pnp4nagios, rrdtool directly etc. Nagios doesn't care/process what your check value is, it only cares about the return code, but you can have eventhandlers that fire based on a warning or critical result. For example, I have an app on a linux box that tends to fail on occasion, the server check has an eventhandler that will ssh to that box and restart the service when it fails rather than annoy me with it.

I integrate nagios and cacti together (cacti for graphing and trending), it works pretty well.

The result codes are:
0 = ok
1 = warning
2 = critical
3 = unknown/unreachable

You can absolutely return any values you want. You can have pass/fail, several steps as you describe, or set those levels at higher numbers. A great example is checking the number of processes on a server - 100 might warn, 200 might critical. From there you can graph that data, as Dave said with RRD.

I use Icinga (a Nagios fork) and it has nice integration with RRDtools. I don't use the flash based GUI, the HTTP based one is also a bit better than plain Nagios and without the issues Flash brings about. One of the nice things is configurations can move back and forth between Icinga and Nagios installs without issue.