Facter via YAML¶

Overview¶

This method of accessing facts is what we recommend for Puppet users, it’s faster and deals better with some strange issues in Facter.

Essentially we use Puppet to dump the data it gathers during normal runs into a YAML file and just use the MCollective YAML fact source to read this data. This avoids an extreme slowdown from Facter running on every mcollective invocation, plus it lets you get any in-scope variables (for example, parameters from your external node classifier) available as mcollective filters for free.

Creating the YAML¶

In your Puppet manifests create a file resource similar to below:

file{"/etc/mcollective/facts.yaml":
   owner    => root,
   group    => root,
   mode     => 400,
   loglevel => debug,  # this is needed to avoid it being logged and reported on every run
   # avoid including highly-dynamic facts as they will cause unnecessary template writes
   content  => inline_template("<%= scope.to_hash.reject { |k,v| k.to_s =~ /(uptime_seconds|timestamp|free)/ }.to_yaml %>")
}

Apply this to all the nodes that run MCollective.

Configure MCollective¶

Add the following lines to server.cfg

factsource = yaml
plugin.yaml = /etc/mcollective/facts.yaml

Verify¶

% mc-inventory some.node
Inventory for some.node:

   Server Statistics:
                      Version: 0.4.10
                   Start Time: Mon Nov 29 16:38:28 +0000 2010
                  Config File: /etc/mcollective/server.cfg
                   Process ID: 5387
               Total Messages: 9254
      Messages Passed Filters: 6443
            Messages Filtered: 2811
                 Replies Sent: 6442
         Total Processor Time: 23.27 seconds
                  System Time: 6.19 seconds



   Facts:
      architecture => i386
      clientcert => some.node
      clientversion => 2.6.3
      country => de