Skip to content
This repository

Cook with Chef without a Chef Server

README.md

LittleChef

With LittleChef you will be able to get started more quickly cooking with Chef, the excellent Configuration Management System.

Overview

You may think of this like a pocket Chef that doesn't need a Chef Server. Just your local kitchen with all your cookbooks, roles data bags and nodes, which will get rsynced to a node each time you start a Chef Solo configuration run with the bundled fix command.

It also adds features to Chef Solo that are currently only available for Chef Server users: data bag search, and node search.

How it all works

It all starts in the kitchen, which you should keep under version control:

  • littlechef.cfg: Configuration, including authentication and run-time options.
  • nodes/: After recipes are run on Nodes, their configuration is stored here in JSON format. You can manually edit them or even add new ones. The name of a node should be its FQDN
  • cookbooks/: This will be your Cookbooks repository
  • site-cookbooks/: Here you can override upstream cookbooks (Opscode's, for example)
  • environments/: Where Chef Environments are defined in JSON
  • roles/: Where Chef Roles are defined in JSON
  • data_bags/: Chef Data Bags. JSON databag items. Search is supported
  • plugins/: Your plugin tasks

Whenever you start a Chef Solo configuration run with the local fix command, all cookbooks, roles and databags are rsynced to the /tmp/chef-solo/ directory, together with the /etc/chef/node.json and /etc/chef/solo.rb files, and chef-solo is executed at the remote node.

The result is that you can configure your nodes exactly when and how you want, all without needing a Chef Server. And all your infrastructure, including your nodes, will be in code, revision controlled.

Environments

Chef 10 does not support Environments in Solo mode, but LittleChef will automatically add a cookbook library that will let you define chef_environment in a role or node.

If you use Chef 11, all environment features except cookbook versioning work out of the box.

Data bag Search

Chef Solo does not support data bag search. To be able to use search, the chef-solo-search library needs to be present in your cookbooks directory.

Most examples in the [chef search documentation][] are possible, including stuff like: search(:users, "married:true AND age:35").

Node Search

Node search is achieved by creating a "node" data bag on the fly for every run, with the data from each node defined in nodes/, but with the attribute values being the result from merging cookbook, node and role attributes, following the standard Chef attribute preference rules. Some automatic attributes are also added.

munin_servers = search(:node, "role:#{node['munin']['server_role']} AND chef_environment:#{node.chef_environment}")

Logs

Chef Solo output for a configuration run will be found at the node's /var/log/chef/solo.log, and the previous configuration run will be moved to solo.log.1.

metadata.rb and ruby roles

LittleChef depends on the JSON versions of the cookbook metadata and roles to properly merge attributes. You can still use the ruby versions, and generate the JSON versions when you make changes. If you have knife locally installed, it will even be done automatically on every run if a changed metadata.rb is detected. Ruby roles are not yet automatically converted, but an implementation is planned.

Plugins

You can define your own LittleChef tasks as Python plugin modules. They should be located in the plugins directory. The filename will be the plugin name and the module docstring the description. Each plugin should define an execute funtion, which will then be executed when applying a plugin on a node (the Cooking section describes how to run a plugins).

You can find example plugins in the repository plugins directory

Getting Big

Beyond a small number of nodes it becomes more and more difficult to keep track of everything. To solve that problem we created Kitchen, an incredibly useful dashboard that provies an overview of all nodes and a graphical visualization of role dependencies.

Installation

Desktop support

Tested on all major operating systems: Linux, Mac OS X, Windows and FreeBSD

Requirements

  • Python 2.6+
  • Fabric 1.5+

The best way to install LittleChef is using pip. Required packages are installed by typing:

$ sudo apt-get install python-pip python-dev

for Debian and Ubuntu, or

$ yum install python-pip python-devel

for RHEL and CentOS.

Installation

You can install LittleChef directly from the PyPI:

$ pip install littlechef

Usage

Disclaimer

Careful what you do with your nodes!:

A certain famous Chef: What do I always say? Anyone can cook.
LittleChef: Yeah. Anyone can, that doesn't mean that anyone should.

Local Setup

To create a new kitchen, type:

$ fix new_kitchen

This will create a few files and directories inside the current directory which LittleChef needs to be able to cook: littlechef.cfg, environments, roles/, data_bags/, nodes/, cookbooks/ and site-cookbooks/. You can create and have as many kitchens as you like on your computer.

Authentication

To be able to issue commands to remote nodes, you need to enter a user and a password with sudo rights. new_kitchen will have created a file named littlechef.cfg. You can edit it now to enter needed authentication data. There are several possibilities:

  • username and password
  • username, password and keypair-file
  • A reference to an ssh-config file

The last one allows the most flexibility, as it allows you to define different usernames, passwords and/or keypair-files per hostname. LittleChef will look at ~/.ssh/config by default, but you can always specify another path in littlechef.cfg:

[userinfo]
user = myusername
password = mypassword
ssh-config = /path/to/config/file

An example ~/.ssh/config file:

Host www.cooldomain.com
    HostName www.cooldomain.com
    IdentityFile ~/.ssh/prod_rsa
    User produser
Host *.devdomain.com
    IdentityFile ~/.ssh/dev_rsa
    User devuser

Other Configuration Options

You can also optionally override the directory being used on the nodes to sync your kitchen to:

[kitchen]
node_work_path = /tmp/chef-solo

If you're using encrypted data bags you can specify a path for the encrypted_data_bag_secret file:

[userinfo]
encrypted_data_bag_secret = ~/path/to/encrypted_data_bag_secret

This will put the encrypted_data_bag_secret in /etc/chef/encrypted_data_bag_secret with permissions root:root with perms 0600. Chef-solo will automatically use it wherever you use Chef::EncryptedDataBagItem.load in your recipes. It will also remove the /etc/chef/encrypted_data_bag_secret file from the node at the end of the run.

If your nodes are not directly accessible, you might want to specify a gateway host. The fix command will connect to the host specified and issue all following connections from this host. All ssh communication will be tunneled through this gateway connection. This can be used if your nodes are behind a firewall and only one host is accessible from your current network location.

[connection]
gateway = hub.example.com

After issuing a fix command, this will connect to hub.example.com. All further node connections will be done from hub.example.com.

The sync-packages section allows you to define remote and local directories, which will then be synchronized at every run.

[sync-packages]
dest-dir = /srv/packages
local-dir = ./packages

Deploying chef-solo

For convenience, there is a command that allows you to deploy chef-solo to a node.

Omnibus method

The best way is to use the omnibus method getchef: fix node:MYNODE deploy_chef:method=omnibus,version=11.12

Other methods

The default installation method: fix node:MYNODE deploy_chef uses the packages from the Opscode repository, which are no longer updated, so its use is no longer recommended. LittleChef will try to autodetect the distro type and version of that node, and will use the appropriate installation method and packages. LittleChef 2.x will use the omnibus method as default instead.

You can also install Chef Solo with gems and/or without asking for confirmation:
fix node:MYNODE deploy_chef:gems=yes,ask=no

Currently supported Linux distributions include Ubuntu, Debian, CentOS, RHEL, Scientific Linux, Gentoo, and Arch Linux.

When using the Debian repository, you need to take into account that Opscode has separated Chef versions in different repos. Current default is Chef 0.10, but you can install Chef 0.9 by typing: fix node:MYNODE deploy_chef:version=0.9

Also, if you still want to keep the chef-client around in debian, use the stop_client option: fix node:MYNODE deploy_chef:stop_client=no

Note that if you already have Chef Solo installed on your nodes, you won't need this. Also, if you previously installed Chef using the Gem procedure, please don't use the deploy_chef package installation method, removing the gem first might be a good idea.

Cooking

Note: Don't cook outside of a kitchen!

List of commands:

  • fix -v: Shows the version number
  • fix -l: Show a list of all available orders
  • fix -y: Automatic yes to prompts; assume "yes" as answer to all prompts and run non-interactively
  • fix node:MYNODE role:MYROLE: Full chef-solo configuration run on the given node. Uses ssh_config hostname directives
  • fix node:MYNODE1,MYNODE2: Configures several nodes. For all "fix node" commands, nodes with the attribute dummy set to true will not be configured and will be skipped
  • fix node:all: It will apply all roles, recipes and attributes defined for each and every node in nodes/
  • fix --env=MYENV node:all: Configures all nodes which have the attribute chef_environment set to MYENV
  • fix node:MYNODE role:MYROLE: Apply the given role on the given nodes
  • fix node:MYNODE recipe:MYRECIPE: Apply the given recipe on the given nodes
  • fix nodes_with_role:ROLE1: Configures all nodes which have the given role in their run_list
  • fix nodes_with_role:ROL*: Configures all nodes which have at least one role which starts with 'ROL' in their run_list
  • fix nodes_with_recipe:MYRECIPE: Configures all nodes which have the given recipe in their run_list
  • fix nodes_with_tag:MYTAG: Configures all nodes which have the tag MYTAG assigned. --include-guests will also configure guest nodes belonging to each host which has the given tag assigned
  • fix node:MYNODES ssh:"my shell command": Executes the given command on the node
  • fix node:MYNODES plugin:save_ip: Gets the actual IP for this node and saves it in the ipaddress attribute

Options:

  • fix --env=MYENV nodes_with_role:ROLE1: Configures all nodes in the environment MYENV which have a certain role in their run_list.
  • fix --verbose node:MYNODE: Chef 0.10.6 introduced the verbose_logging option. When false, the "processing" messages are not longer shown. That is the new default for LittleChef, so that you now only see what has changed in this configuration run. --verbose switches this back on.
  • fix --debug node:MYNODE: You can start all your commands with fix --debug to see all Chef Solo debugging information. Also, the node file and node databag wont't be deleted from the node, and verbose will also be true
  • fix --no-report node:MYNODE: will prevent the logging of Chef Solo output to /var/log/chef/
  • fix --why-run node:MYNODE: will configure the node in Whyrun mode

Once a node has a config file, the command you will be using most often is fix node:MYNODE, which allows you to repeatedly tweak the recipes and attributes for a node and rerun the configuration.

Configuring nodes in parallel

By default LittleChef configures nodes serially however it can also use Fabric's parallel SSH support to configure multiple nodes in parallel. All commands are supported (node, nodes_with_role, ssh, role, and recipe)

  • fix --concurrency node:NODELIST: will configure multiple nodes in parallel
  • fix --concurrency node:NODELIST ssh:COMMAND: will run an ssh command on multiple nodes in parallel
  • fix --concurrency 5 node:NODELIST: will configure multiple nodes in parallel but limit to 5 connections

Consulting the inventory

  • fix list_nodes: Lists all configured nodes, showing its associated recipes and roles
  • fix list_nodes_detailed: Same as above, but it also shows all attributes
  • fix list_nodes_with_recipe:MYRECIPE: Lists nodes which have associated the recipe MYRECIPE
  • fix list_nodes_with_role:MYROLE: Shows nodes which have associated the role MYROLE
  • fix list_nodes_with_tag:MYTAG: Shows nodes which have the tag MYTAG assigned. --include-guests will also display guest nodes belonging to each host which has the given tag assigned
  • fix list_envs: Lists all available environments
  • fix list_recipes: Lists all available recipes
  • fix list_recipes_detailed: Same as above, but shows description, version, dependencies and attributes
  • fix list_roles: Lists all available roles
  • fix list_roles_detailed: Same as above, but shows description and attributes
  • fix list_plugins: Show a list of available plugins

Using LittleChef as a library

You can import littlechef.py into your own Python project. The following script is equivalent to using the fix orders:

from littlechef import runner as lc
lc.env.user = 'MyUsername'
lc.env.password = 'MyPassword'
lc.env.host_string = 'MyHostnameOrIP'
lc.deploy_chef(gems='yes', ask='no')

lc.recipe('MYRECIPE') #Applies <MYRECIPE> to <MyHostnameOrIP>
lc.node('MyHostnameOrIP') #Applies the saved nodes/MyHostnameOrIP.json configuration

Performance Tips

You can greatly reduce the SSH connection setup time by reusing existing connections. On Unix systems, you can do so by adding the ControlMaster directive to your ssh config:

#~/.ssh/config
Host *
  ControlMaster auto
  ControlPath /tmp/ssh-%r@%h:%p

Other tutorial material

Getting help

For help regarding the use of LittleChef, or to share any ideas or suggestions you may have, please post on LittleChef's discussion group

Reporting bugs

Build Status

If you find bugs please report it on https://github.com/tobami/littlechef/issues

Happy cooking!

Something went wrong with that request. Please try again.