Dynamic commands, services, and NRPE associations

[thenewwazoo]

This is the first of two forthcoming PRs relating to dynamic Nagios configuration using Salt.

This PR adds the capability to define a Nagios service check in one place in the pillar data, and for both the NRPE-daemon-running minion and the Nagios-running minion to emit appropriate configurations for each. The data added to pillar.example illustrates how to use this data. As the author, it's hard to judge how straightforward this is; please let me know if actual documentation is necessary.

There is one breaking change: I changed the name of the nagios.nrpe.checks dict (used to embed command[...]=... lines directly into nrpe.cfg) to be nagios.nrpe.nrpe_commands. This is to reduce confusion about where checks (in the broader sense) are defined vs simpler command specifications.

This PR also includes fixes in pillar.example relating to use of the nagios.lookup dict (and the key names defined therein).

Still TODO is porting the dynamic NRPE configuration emitter states to non-Debian distributions (read: distros which don't do /etc/nagios/nrpe.d), and validation that /usr/lib/nagios/plugins/ is a sane default for nrpe.plugin_dir (see map.jinja).

[puneetk]

In the next pull, can you please update the readme to doc the breaking change and possibly use salt'grains.get' and salt'pillar.get'

[wwentland]

Oh man .. This breaks backwards compatibility as checks is being replaced with nrpe_commands in the pillar and was even mentioned in the discussion. Changes to the pillar datastructure shouldn't be done if there isn't a very good reason as many people might have pushed changes that caused them to remove all nrpe commands from their respective config.

[wwentland]

I am in favour of reverting this PR as it causes a lot of issues and forces users to adapt their pillars. What do you guys think?

[gravyboat]

@BABILEN I'm fine with reverting it since I don't think the issues that were originally brought up were fixed. Just keep in mind it's been around for almost a year now. Is it going to be less work to fix it?

[thenewwazoo]

Sorry, I thought it was pretty clear when I wrote "There is one breaking change..." The reason for renaming checks to nrpe_commands is because a nrpe command isn't a check, and was colliding with the (properly named) checks object in the nagios config.

I haven't been following issues, but I've gone through anything linked. What issues remain unresolved? I'm happy to chime in if there's any question.

[wwentland]

Sure, fixing the issues would be much preferred. Let's see if fixing #28 sorts out the last issue.