Skip to content
This repository

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
Fetching contributors…

Octocat-spinner-32-eaf2f5

Cannot retrieve contributors at this time

file 358 lines (221 sloc) 9.85 kb

NAME

Pantry - Configuration management tool for chef-solo

VERSION

version 0.005

SYNOPSIS

  $ mkdir my-project
  $ cd my-project
  $ pantry init
  $ pantry create node foo.example.com
  $ pantry list nodes
  $ pantry apply node foo.example.com --recipe nginx
  $ pantry apply node foo.example.com --default nginx.port=80
  $ pantry sync node foo.example.com

DESCRIPTION

pantry is a utility to make it easier to manage a collection of computers with the configuration management tool chef-solo http://wiki.opscode.com/display/chef/Chef+Solo

USAGE

Arguments to the pantry command line tool follow a regular structure:

  $ pantry VERB [[NOUN] [ARGUMENTS...]]

See the following sections for details and examples by topic.

Pantry setup and introspection

init

  $ pantry init

This initializes a pantry in the current directory. Currently, it just creates some directories for use storing cookbooks, node data, etc.

list

  $ pantry list nodes
  $ pantry list roles

Prints to STDOUT a list of nodes or roles managed within the pantry.

Managing nodes

In this section, when a node NAME is required, the name is expected to be a valid DNS name or IP address. The name will be converted to lowercase for consistency. When referring to an existing node, you may often abbreviate it to a unique prefix, e.g. "foo" for "foo.example.com".

Also, whenever a command takes a single 'node NAME' target, you may give a single dash ('-') as the NAME and the command will be run against a list of nodes read from STDIN.

You can combine this with the pantry list command to do batch operations. For example, to sync all nodes:

  $ pantry list nodes | pantry sync node -

create

  $ pantry create node NAME

Creates a node configuration file for the given NAME.

rename

  $ pantry rename node NAME DESTINATION

Renames a node to a new name. The old node data file is renamed. The NAME must exist.

delete

  $ pantry delete node NAME

Deletes a node. The NAME must exist. Unless the --force or -f options are given, the user will be prompted to confirm deletion.

show

  $ pantry show node NAME

Prints to STDOUT the JSON data for the given NAME.

apply

  $ pantry apply node NAME --recipe nginx --role mail --default nginx.port=80

Applies recipes, roles or attributes to the given NAME.

To apply a role to the node's run_list, specify --role role or -R role. May be specified multiple times to apply more than one role. Roles will be appended to the run_list before after any existing entries but before any recipes specified in the same command.

To apply a recipe to the node's run_list, specify --recipe RECIPE or -r RECIPE. May be specified multiple times to apply more than one recipe.

To apply an attribute to the node, specify --default KEY=VALUE or -d KEY=VALUE. If the KEY has components separated by periods (.), they will be interpreted as subkeys of a multi-level hash. For example:

  $ pantry apply node NAME -d nginx.port=80

will be added to the node's data structure like this:

  {
    ... # other node data
    nginx => {
      port => 80
    }
  }

If the VALUE contains commas, the value will be split and serialized as an array data structure. For example:

  $ pantry apply node NAME -d nginx.port=80,8080

will be added to the node's data structure like this:

  {
    ... # other node data
    nginx => {
      port => [80, 8080]
    }
  }

Both KEY and VALUE support periods and commas (respectively) to be escaped by a backslash.

N.B. While the term --default is used for command line consistency, attributes set on nodes actually have what Chef terms "normal" precedence.

strip

  $ pantry strip node NAME --recipe nginx --role mail --default nginx.port

Strips recipes, roles or attributes from the given NAME.

To strip a role to the node's run_list, specify --role role or -R role. May be specified multiple times to strip more than one role.

To strip a recipe to the node's run_list, specify --recipe RECIPE or -r RECIPE. May be specified multiple times to strip more than one recipe.

To strip an attribute from the node, specify --default KEY or -d KEY. The KEY parameter is interpreted and may be escaped just like in apply, above.

sync

  $ pantry sync node NAME

Copies cookbooks and configuration data to the NAME node and invokes chef-solo via ssh to start a configuration run. After configuration, the latest run-report for the node is updated in the 'reports' directory of the pantry.

edit

  $ pantry edit node NAME

Invokes the editor given by the environment variable EDITOR on the configuration file for the name node.

The resulting file must be valid JSON in a form acceptable to Chef. Generally, you should use the apply or strip commands instead of editing the node file directly.

Managing roles

In this section, when a role NAME is required, any name without whitespace is acceptable. The name will be converted to lowercase for consistency. When referring to an existing role, you may often abbreviate it to a unique prefix, e.g. "web" for "webserver".

Also, whenever a command takes a single 'node NAME' target, you may give a single dash ('-') as the NAME and the command will be run against a list of roles read from STDIN.

You can combine this with the pantry list command to do batch operations. For example, to add a recipe to all roles:

  $ pantry list roles | pantry apply role - --recipe ntp

create, rename, delete, show and edit

These commands work the same as they do for nodes. The difference is that you must specify the 'role' type:

  $ pantry create role web
  $ pantry show role web

apply and strip

The apply and strip commands have slight differences, as roles have two kinds of attributes, "default attributes" (--default or -d) and "override attributes" (--override or -O), with slightly different precedence.

  $ pantry apply role NAME -d nginx.user=nobody -O nginx.port=80
  $ pantry strip role NAME -d nginx.user -O nginx.port

The --recipe (-r) and --role (-R) arguments work the same as for nodes. Note that roles can have other roles in their run_list.

When Chef merges attribute, the role default attribute has the lower precedence than node attributes. Override attributes have higher precedence than node attributes. Yes, this is a gross simplification of how Chef does it. See Chef docs for more: http://wiki.opscode.com/display/chef/Attributes

Getting help

commands

  $ pantry commands

This gives a list of all pantry commands with a short description of each.

help

  $ pantry help COMMAND

This gives some detailed help for a command, including the options and arguments that may be used.

AUTHENTICATION

pantry relies on OpenSSH for secure communications with managed nodes, but does not manage keys itself. Instead, it expects the user to manage keys using standard OpenSSH configuration and tools.

The user should specify SSH private keys to use in the ssh config file. One approach would be to use the IdentityFile with a host-name wildcard:

  IdentityFile ~/.ssh/identities/id_dsa_%h

This would allow a directory of host-specific identities (which could all be symlinks to a master key). Another alternative might be to create a master key for each environment:

  IdentityFile ~/.ssh/id_dsa_dev
  IdentityFile ~/.ssh/id_dsa_test
  IdentityFile ~/.ssh/id_dsa_prod

pantry also assumes that the user will unlock keys using ssh-agent. For example, assuming that ssh-agent has not already been invoked by a graphical shell session, it can be started with a subshell of a terminal:

  $ ssh-agent $SHELL

Then private keys can be unlocked in advance of running pantry using ssh-add:

  $ ssh-add ~/.ssh/id_dsa_test
  $ pantry ...

See the documentation for ssh-add for control over how long keys stay unlocked.

ROADMAP

In the future, I hope to extend pantry to support some or all of the following:

  • Chef role creation and application
  • environments
  • tagging nodes
  • searching nodes based on configuration
  • data bags
  • cookbook download from Opscode community repository
  • bootstrapping Chef over ssh

If you are interested in contributing features or bug fixes, please let me know!

SEE ALSO

Inspiration for this tool came from similar chef-solo management tools. In addition to being implemented in different languages, each approaches the problem in slightly different ways, neither of which fit my priorities. Nevertheless, if you use chef-solo, you might consider them as well:

SUPPORT

Bugs / Feature Requests

Please report any bugs or feature requests through the issue tracker at https://github.com/dagolden/pantry/issues. You will be notified automatically of any progress on your issue.

Source Code

This is open source software. The code repository is available for public review and contribution under the terms of the license.

https://github.com/dagolden/pantry

  git clone https://github.com/dagolden/pantry.git

AUTHOR

David Golden <dagolden@cpan.org>

COPYRIGHT AND LICENSE

This software is Copyright (c) 2011 by David Golden.

This is free software, licensed under:

  The Apache License, Version 2.0, January 2004
Something went wrong with that request. Please try again.