PE 2.7.0: redeclaring pe-httpd service

[carthik]

Using the puppetdb module on a PE-2.7.0 node causes the following error:

err: Could not retrieve catalog from remote server: Error 400 on SERVER: Duplicate declaration: Service[pe-httpd] is already declared in file /etc/puppetlabs/puppet/modules/puppetdb/manifests/master/config.pp at line 116; cannot redeclare at /opt/puppet/share/puppet/modules/request_manager/manifests/init.pp:23 on node classroom.puppetlabs.vm

request_manager is a part of Puppet Enterprise, and adds the new "node requests" certificate management feature.

[cprice404]

So, the PuppetDB module is declaring the service within an "if !defined" block, while the request_manager module does not appear to be doing so. I'm not sure what "best practices" are here, but it seems like there are 2 possible resolutions:

1. Make sure that the request_manager module is doing the same so that they don't compete with one another, or
2. Move the declaration of the pe-httpd service into its own module/class, as per @haus suggestion, and then have both of these two modules reference that class.

I'll investigate a bit further to see if option 1 would be an acceptable short-term solution.

[cprice404]

Reviewing option 1 from my previous commit message; it turns out that the module we're conflicting with (PE request_manager) specifies an extra parameter in the service resource declaration that we are not specifying in the PuppetDB module. Thus, we are going to end up with some weird issues based on the order that the classes are applied if we go with that proposal... so... that's bad. I'm going to open an issue on that module and also on the PE sprintly board to try to make sure this gets the proper attention at some point.

[binford2k]

My workaround for the time being was to set restart_puppet => false and then restart puppet by hand post installation.

class { 'puppetdb::master::config':
  puppetdb_server => 'puppetdb',
  restart_puppet  => false,
}

[hunner]

The pe-httpd service should definitely be managed by a module that is neither puppetdb or request_manager (because they are not the same application). That would be your option 2.

In request_manager you mentioned that it added a second attribute. It is a subscribe so you can simply do File['/etc/puppetlabs/puppet/auth.conf'] ~> Service['pe-httpd'] or perhaps better yet File['/etc/puppetlabs/puppet/auth.conf'] ~> Class['pe_apache::service'] as is common practice (assuming the name of the module that takes care of configuring pe-httpd is called pe_apache).

Theoretically, the puppetlabs-apache module could be used by pe_apache to manage much of the configuration. And it it can't then it should be patched to be that flexible ;).

[cprice404]

Thanks @hunner , that's helpful.

[h0tw1r3]

PE no longer uses this module to manage postgresql.