Using MCollective to run Puppet

clusterconfiguration-managementmcollectivepuppet

I have to administrate a big cluster of servers in the cloud and now I want to automate the installation process of these machines. Therefore I use puppet as configuration management system.

I want to push configurations from the master node to the cluster nodes very fast and heard, that MCollective can do this. But I don't know how the interaction and setup between MCollective, RabbitMQ and Puppet has to be.

Is there a good tutorial about that on the Internet or can someone give me a short intro about this?

This would be great!

Thanks,
Markus

Best Answer

The easiest integration between puppet and mcollective is with the Puppetd Agent and, optionally, puppet commander -

http://projects.puppetlabs.com/projects/mcollective-plugins/wiki/AgentPuppetd

http://projects.puppetlabs.com/projects/mcollective-plugins/wiki/ToolPuppetcommander

Don't use rabbitmq, especially if you have any significant scale. ActiveMQ is the best supported, best documented queuing system for use with mcollective.

To get started check out RI Pienaar's screencasts page -- the docs are pretty good but the screencasts will get you a handle on the concepts much faster. I especially recommend the Message Flow, Components, and Middleware screencast.

Related Solutions

Ubuntu – How to install puppet automatically on image startup

Puppet client creates new certificates on start if they are missing. So, you just need to install puppet (with puppet master hostname configured) and ensure that its certificate is not included with image (by removing it right before making an image).

On bootup of freshly cloned machine, puppet would generate new certificates and will make connection attempt to master. And you'll see it in output of puppetca --list

Using puppet dashboard as an enc

I managed to fix this. Once all the config was correct, things all worked ok.

The main areas were - the externode_node needs to be on all puppet masters and the puppet dashboard server. The cert locations should be the ones that are created by the rake tools when enabling https on dashboard. You can see these in the settings.yml file on the dashboard server. Make sure in the DASHBOARD_URL you are using the common-name in the cert created by the rake tool, in most cases dashboard. You may need to set up a c-name for dashboard or an A-record if you so desire. Make sure the working external_node script is copied to all puppet master servers and is the same. The location I used was /usr/share/puppet-dashboard/bin/external_node. Make sure the URL is https:// dashboard or the cn of the dashboard cert. Otherwise you will get an SSL name does not match certname error

The puppet.conf file on each master has 2 lines to enable the ENC. This is as follows:

node_terminus = exec
external_nodes = /usr/bin/env PUPPET_DASHBOARD_URL=https:// dashboard/usr/share/puppet-dashboard/bin/external_node

See the dashboard name - that is the same as the CN name in the dashboard cert. Make sure you restart httpd on each puppet master. If you are still reading this, good luck!

Best Answer

Related Solutions

Ubuntu – How to install puppet automatically on image startup

Using puppet dashboard as an enc

Related Topic