Home / server / Questions / 340603
In Process
Andrew McCloud
Asked: 2011-12-14 11:49:50 +0800 CST

Bare minimal Chef provisioning and deployment?

  • 772

I've read the documentation on Chef twice over. I still can't wrap my head around it's concept because they skip but fundamentals and jump to complex deployments with chef-server.

Using chef-solo and possibly knife, is there a simple way to provision a server and deploy?

I may be wrong, but it seems like with my cookbooks prepped, this should be very simple.

knife rackspace server create --flavor 1 --image 112

That provisions my server. I can optionally pass --run-list "recipe[mything]", but how do my cookbooks in ~/my_cookbooks actually get on the server? Do I have to manually transfer them? That seems counterproductive.

chef chef-solo knife

5 Answers

  1. If you use chef-solo, you don't get to use knife. Knife is API client for the chef-server, with some extra utility sugar (like knife rackspace server create you've mentioned).

    To configure server with chef-solo, you should copy your chef repo to the server, and run chef-solo there over ssh. There is no ready-made script or knife plugin that I know of that would do it automatically.

    Command knife rackspace server create creates new Rackspace server, and then initializes it for chef-server that knife knows of by calling knife bootstrap. It won't work with chef-solo easily.

    Technically, knife bootstrap, and thus knife rackspace server create, can be coerced to work with chef-solo by writing a custom bootstrap template that, instead of initializing chef-client, would download your chef repository and run chef-solo - see knife bootstrap --help, its wiki page, or source for details. You can see example templates for installing chef-client here. This is an advanced feature, though, and it's not very well documented.

    If you don't want to handle complexity of installing and managing your own chef-server, you can use free layer of Opscode's Hosted Chef, which is Chef-server SAAS offering and is free up to three nodes. I'd recommend starting any serious work with server anyway - chef-solo is as good as a decent bootstrap shell script, no more, and you're missing out on many important/interesting features like search and data bags, which allow you to configure your servers in a data-driven way.

    • 6

  2. Check out the knife-solo plugin, which can automatically install chef-solo on your remote server, upload your cookbooks onto it, and then run chef-solo. It basically automates what other folks who've answered this question have suggested doing.

    • 5

  3. Before each chef-solo run, the cookbooks should be present on that target machine, either by transferring it (via ftp/scp) or pointing the cookbook_path to a network share .

    If you want the cookbooks to be automatically downloaded, you would need to run a Chef Server. Whether you want to run your own chef server, or use a hosted one from OpsCode is up to you, but it is needed.

    • 2

  4. Barest, most basic way to get going with chef-solo.

    Examples are contrived, and you should modify them.

    More information about Chef Solo:

    While Chef Solo is useful, it is really a limited way to use Chef. It doesn't expose information about the node outside of the node itself, so it cannot be used for dynamic discovery or data-driven infrastructure management nearly as easily as with Chef Server.

    • 1

  5. There's a git project for that ;)

    The knife-solo git project should enable you to run knife in conjunction with chef-solo and allow you to do remote provisioning from your local dev environment to remote servers like this once you've got it configured locally and installed:

    knife solo cook [email protected]

    • 0