Some time ago I posted a blog post and screencast about my middleware solution to concurrent systems administration, I had a lot of interest in this and in the end relented and released the thing as Open Source. You should go watch the screencast to really grasp what is going on and what this achieves.
I asked around on Twitter for name suggestions and got many great suggestions, in the end I settled on mcollective short for The Marionette Collective, yeah it sounds like an Emo band, I don’t care. It’ll be easy to google and I got a matching .org domain, that helps over certain other ungoogleable project names!
The code for the core service is released under the Apache License version 2 and is available at google code: http://marionette-collective.org/
There’s a wiki getting started guide and also a quick guide on writing agents and CLI tools to talk to it.
By default the code includes the following plugins:
- Security by means of a Pre Shared Key
- Communications with Stomp compatible servers
- Facts about machines via a simple YAML file
- Discovery
You can drop in replacement agents for all of the above and perhaps write something to use full SSL encryption etc. No actual agents to do any real work is included in the core Apache Licensed code.
A second project was started where I’m dropping a bunch of my own agents, the projects are separate because the agents might have different licensing from the core app server, for example there’s an agent to use Facter from Reductive Labs for facts but this means the code has to be GPL.
Agents available right now lets you use facter for facts, manage services using the puppet providers and also do distributed url benchmarks, check out the plugins project: http://code.google.com/p/mcollective-plugins/
I’d encourage others to put agent on Github or wherever and to announce it on the users list or just to drop me a mail and I’ll link to it from the project wiki – you can also join one of the projects and just commit your code yourself.
It’s still early days – I was accused of being too perfectionist in how I like to release code, so this is very much an early and often approach to releasing! The entire code base is about 3 weeks old and I spent mostly some free time hacking it up, so there’s much improvement to be made still, I do however use a version of it in production and find it very stable and reliable so far.
I am looking for early testers to give me feedback about the code, structure of the project etc. If you’re stuck grab me on freenode my nick is Volcane and I’ll see if I can help you get it going.
