[chef] Re: Chef on existing infrastructure/nodes

["John E. Vincent (lusis)" < >]

I tend to thing of the approach like this game:

http://s3itch.lusis.org/sliding-puzzle-20130423-094943.jpg

In almost every company where I've introduced Chef or Puppet, the first thing we did was:

- Automate some basic mundane configuration on existing systems (ntp, ssh key distribution, whatever)

- Find at least ONE blank system (virtualized, physical - it doesn't matter)

- Start developing and rebuilding the system with cookbooks/modules

- Do that one system OVER AND OVER AND OVER AND OVER until you can go soup to nuts (bootstrap to in-server)

- Move it into production and terminate the old system

- Use the newly decommissioned system to repeat the process with another component

This will take time. Lots of time but the end result is that you've rebuilt your entire infra in something repeatable. At the last company it took me about 3 months (we were all on AWS so I had free resources). At the current company it took 6-8 months as we were running on cloudstack privately with non-infinite resources.

On Tue, Apr 23, 2013 at 9:44 AM, Mike Dillion < " target="_blank"> > wrote:

Hey y'all,

I heard something on a Puppet Labs podcast* recently that gave me a pause for though: developing cookbooks (manifests in Puppet?) against production machines as a way of migrating to a coded infrastructure.

i.e. instead of starting from scratch and attempting to duplicate current infrastructure, slowly adding pieces (automating processes) to cookbooks in order to gain immediate benefits.

Are there any patterns for this? Does anyone have any experience, good or bad? I thought it was interesting at the least.

Cheers,

Mike Dillion

* https://puppetlabs.com/blog/devops-culture-podcast/