As of R10K 1.0.0 the purgedirs functionality and configuration options
have been removed. This module continues to configure the no longer
acknowledged configuration options.

This commit removes support for the purgedirs configuration option.

…the pe_gem provider is used. This prevents the puppet run from failing on foss puppet with strict variables enabled

… rather than test for pe_gem being specified as the the provider for the package

…or true, to account for stringified facts, as the getvar() approach still throws an error under future parser and puppet 4 on the foss version.

  - Tested with OpenBSD -current, where there is a r10k package available
  - Ruby binaries on OpenBSD are prefixed with the Ruby version they are built
    for, therefore r10k is called r10k21, since that is the version Puppet
    and r10k relies on
    - that would also be the case if r10k would be installed with the 'gem'
      package provider
  - the service name of mcollective is called 'mcollectived' on OpenBSD

Because of the last two mentioned points, added an osfamily dependent
 r10k_binary variable, as well as making the mc_service_name variable
osfamily dependent too.

What I tested was managing the installation of package and configuration
file /etc/r10k.yaml as well as the prerun script added to /etc/puppet/puppet.conf
file. Pre-run script comes in handy, since my modules/hieradata is on
a ram disk, and on boot, it auto-fills itself before a first puppet agent
run on the master.

Not tested is the mcollective integration, since I first have to
clean up my messed up mcollective setup. That's something that
would come later.

Further added lots of specs tests with regard to OpenBSD. Didn't specified
a version, since the install via packages is only since AFTER OpenBSD 5.7,
so its 5.7-current, and I didn't really wanted to base specs on such a fragile
version number, so I better left it off.

* Ruby 1.8.7 does not like the trailing ','
* Switch the path to ruby binary

If enable_ssl is false, the script still tries to read the
public/private_key_path, and completely fails if the files are missing
or invalid.

This fixes the script to NOT pass those parameters (or try to read the
files) unless enable_ssl is true.

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

It should be possible to run the webhook as an unprivileged user (eg
nobody) and only invoke r10k/mco via sudo.

sudoers examples:
nobody ALL=(root) NOPASSWD: r10k deploy environment *, r10k deploy module *
nobody ALL=(peadmin) NOPASSWD: mco r10k deploy *, mco r10k deploy_module *

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Update stash_mco.rb

Add support for OpenBSD, some noteworthy points to mention:

Fix non-ssl, abilliy to customize templates, and run with less privileges (but call r10k/mco via sudo)

…d template.

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

…VerifyClient parameter

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Gentoo & Pidfile/Init fixes

Fix a few typos.

add @ghoneycutt 's comment

Signed-off-by: Robin H. Johnson <robbat2@gentoo.org>

Actually run as non-root, --user is for matching not chuid.

* The parameter is `git_ssl_verify` b
* But the value it's setting is the environment variable `GIT_SSL_NO_VERIFY`
* So git_ssl_verify => 'true' is actually saying "Make sure SSL is not verified"
* Better to rename it to `$git_no_ssl_verify` and add a deprecation to the old parameter
* Followed by removal in a major bump

Closes #165

* Both checks are happy path for now
* But in a major bump, $git_no_verify should either be removed as a parameter, or given a specific error to say "Don't use this anymore"

When use_mcollective is set to false in the webhook.yaml, then avoid including
the mcollective libraries

Also, remove webhook::git_server as it seems useless

Also, drop the webhook_ prefix

The systemd unit files do have status

Fix for #165

webhook: include mcollective optionally

Use the webhook's systemd service file for openSUSE >= 12.1 and SLE12

…atus

Remove the hardcoded hasstatus => false from the webhook service

Add SUSE support in the webhook init file

Move the $webhook::{user,group} definition to params.pp

[Deprecation] Remove purgedirs support

Add a $manage_packages switch on the webhook class that enables the
installation of the webhook needed packages.
Move the installation of the packages to a r10k::webhook::package class
The reason is that we can skip installing the gems, and instead install distro
packages.

- Type=forking as webrick fails to start otherwise
- PIDFile=$pidfile otherwise it fails to auto-detect it, and fails to stop the
  service
- Removing the rest of the variables as they are not really needed, or their
  default values are sufficient

Previously, the r10k run was logged in a separate file. Thus, the webhook
returned always success if the r10k itself failed for some reason to run
properly. This resulted in getting false positives in the CI job.

With this commit the r10k stdout and stderr is sent to the web server, so that
we can get a proper failure message in the json output