With LittleChef you will be able to get started more quickly cooking with Chef, the excellent Configuration Management System.
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.
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 FQDNcookbooks/
: This will be your Cookbooks repositorysite-cookbooks/
: Here you can override upstream cookbooks (Opscode's, for example)environments/
: Where Chef Environments are defined in JSONroles/
: Where Chef Roles are defined in JSONdata_bags/
: Chef Data Bags. JSON databag items. Search is supportedplugins/
: 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.
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.
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 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}")
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
.
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.
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 function, 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
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.
Tested on all major operating systems: Linux, Mac OS X, Windows and FreeBSD
- 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.
You can install LittleChef directly from the PyPI:
$ pip install littlechef
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.
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.
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
littlechef supports Berkshelf. If given file exists littlechef will execute berks vendor berksfile-cookbooks-directory, if use do not provide berksfile_cookbooks_directory then random directory in tmp is used. If user want's to upload some cookbooks which are not tracked in Berskfile, they have to place them to cookbooks directory.
[kitchen]
berksfile = Berksfile
berksfile_cookbooks_directory = berks-cookbooks
You can also optionally override the directory being used on the nodes to sync your kitchen to:
[kitchen]
node_work_path = /tmp/chef-solo
You can use encrypted data bags. Create secret keys, Use knife-solo_data_bag Gem to create encrypted data bags, and 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 you are not comfortable about leaving data bags in the remote node, which will leak sensitive information, you can tell littlechef to delete them all at the end of the run:
[userinfo]
remove_data_bags=true
[userinfo]
autodeploy_chef=true
if set to true, a check will be performed before each deployment for chef-solo, and if it is not present it will be installed using the omnibus method.
[connection]
gateway = hub.example.com
If you want to use http/https proxy with chef_solo run. You have to add following entries to config file. They will create solo.rb config file with http/https proxy configured.
[connection]
http_proxy = "http://172.14.1.12:8888"
https_proxy = "http://172.14.1.12:8888"
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
For convenience, there is a command that allows you to deploy chef-solo to a node.
The best way is to use the omnibus method getchef:
fix node:MYNODE deploy_chef:version=11.12
You can also install Chef Solo without asking for confirmation:
fix node:MYNODE deploy_chef:ask=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 any other procedure, please don't use the deploy_chef installation method, removing chef first might be a good idea.
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.
To be able to properly connect to hub.example.com same host should be defined in ssh-config. Because final connection is made from a hub.example.com any required ssh keys must be available on it or user needs to setup ssh-agent forwarding.
Note: Don't cook outside of a kitchen!
List of commands:
fix -v
: Shows the version numberfix -l
: Show a list of all available ordersfix -y
: Automatic yes to prompts; assume "yes" as answer to all prompts and run non-interactivelyfix node:MYNODE role:MYROLE
: Full chef-solo configuration run on the given node. Uses ssh_config hostname directivesfix node:MYNODE1,MYNODE2
: Configures several nodes. For all "fix node" commands, nodes with the attributedummy
set totrue
will not be configured and will be skippedfix node:all
: It will apply all roles, recipes and attributes defined for each and every node innodes/
fix --env=MYENV node:all
: Configures all nodes which have the attributechef_environment
set toMYENV
fix node:MYNODE role:MYROLE
: Apply the given role on the given nodesfix node:MYNODE recipe:MYRECIPE
: Apply the given recipe on the given nodesfix nodes_with_role:ROLE1
: Configures all nodes which have the given role in their run_listfix nodes_with_role:ROL*
: Configures all nodes which have at least one role which starts with 'ROL' in their run_listfix nodes_with_recipe:MYRECIPE
: Configures all nodes which have the given recipe in their run_listfix nodes_with_tag:MYTAG
: Configures all nodes which have the tagMYTAG
assigned.--include-guests
will also configure guest nodes belonging to each host which has the given tag assignedfix node:MYNODES ssh:"my shell command"
: Executes the given command on the nodefix node:MYNODES plugin:save_ip
: Gets the actual IP for this node and saves it in theipaddress
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 theverbose_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 withfix --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 truefix --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.
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 parallelfix --concurrency node:NODELIST ssh:COMMAND
: will run an ssh command on multiple nodes in parallelfix --concurrency 5 node:NODELIST
: will configure multiple nodes in parallel but limit to 5 connections
fix list_nodes
: Lists all configured nodes, showing its associated recipes and rolesfix list_nodes_detailed
: Same as above, but it also shows all attributesfix list_nodes_with_recipe:MYRECIPE
: Lists nodes which have associated the recipeMYRECIPE
fix list_nodes_with_role:MYROLE
: Shows nodes which have associated the roleMYROLE
fix list_nodes_with_tag:MYTAG
: Shows nodes which have the tagMYTAG
assigned.--include-guests
will also display guest nodes belonging to each host which has the given tag assignedfix list_envs
: Lists all available environmentsfix list_recipes
: Lists all available recipesfix list_recipes_detailed
: Same as above, but shows description, version, dependencies and attributesfix list_roles
: Lists all available rolesfix list_roles_detailed
: Same as above, but shows description and attributesfix list_plugins
: Show a list of available plugins
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
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
- Automated Deployments with LittleChef, nice introduction to Chef using LittleChef
For help regarding the use of LittleChef, or to share any ideas or suggestions you may have, please post on LittleChef's discussion group
If you find bugs please report it on https://github.com/tobami/littlechef/issues
Happy cooking!