Issue 2160 - Ensure nagios_hostgroup() does not break reprovisioning

See https://issues.adblockplus.org/ticket/2160 for more information --

[Felix Dahlke, 2015-03-18 06:27:10]

http://codereview.adblockplus.org/6304894776508416/diff/5629499534213120/modu...
File modules/nagios/manifests/server.pp (right):

http://codereview.adblockplus.org/6304894776508416/diff/5629499534213120/modu...
modules/nagios/manifests/server.pp:197: command => shellquote('sed', '-i',
That's pretty horrible. Can't we do the same on the data we get from hiera (i.e.
before create_resources above)? I'm a bit rusty with Puppet, so I'm not sure,
but I'd MUCH prefer not messing with the file here.

[mathias, 2015-03-18 11:26:35]

http://codereview.adblockplus.org/6304894776508416/diff/5629499534213120/modu...
File modules/nagios/manifests/server.pp (right):

http://codereview.adblockplus.org/6304894776508416/diff/5629499534213120/modu...
modules/nagios/manifests/server.pp:197: command => shellquote('sed', '-i',
On 2015/03/18 06:27:10, Felix H. Dahlke wrote:
> That's pretty horrible. Can't we do the same on the data we get from hiera
(i.e.
> before create_resources above)? I'm a bit rusty with Puppet, so I'm not sure,
> but I'd MUCH prefer not messing with the file here.

I'd be glad if anybody could point out a better solution, maybe I just don't see
the wood for the trees here. Frankly, like I've indicated in the issue tracker,
it's quite a mess:

- A dummy value does not work because Nagios does not accept any unrecognized
input here
- An empty string with q ('""') results in Nagios complaining about an unknown
host
  named '""'.
- There is no explicit Nagios config keyword for the concept of a value being
absent or empty.
- Some sort of default host (e.g. "localhost") obviously makes no sense either.
- One cannot configure a Puppet exec { .. to rm(1) the file only before it's
  updated.
- Iterating through data is a mess in any Puppet version prior to 3.2 (which
introduces each(..)), simply because there is no such thing. 

The common workaround for the latter point is to write wrapper types and apply
the data as input to them - just as we did above. But in our case, I do not see
any way how that could help us circumventing the issue without creating our own,
e.g. template-based alternative for the builtin (!!) nagios_hostgroup(). That
would also be way more code and thus way more prone to error.

Another solution you may like more is just deleting the file at every run. That
of course triggers every run to be seen as an update here, but we've done
similar workarounds before. I actually prefer my sed(1) solution because I do
trust in my regex- and shell-foo.. ;-)

We can also remove this hack, document the issue and maintain the host group
associations manually. But that's basically the same as we did before and makes
even less sense.

[mathias, 2015-03-18 14:41:22]

Never mind, I finally found an easy hack. It's even simpler than all the ones I
found in any thread regarding this bug - but not quite obvious. What about
starting to actually discuss migrating to Puppet 3? Like e.g. Q4 this year.. :-)

[Felix Dahlke, 2015-03-18 15:54:18]

LGTM :D

Yeah, let's move to Puppet 3 :)

[Wladimir Palant, 2015-03-18 16:33:49]

LGTM