SYNOPSIS
shef [named configuration] (options)- -S, --server CHEF_SERVER_URL
- The chef server URL
- -z, --client
- chef-client mode
- -c, --config CONFIG
- The configuration file to use
- -j, --json-attributes JSON_ATTRIBS
- Load attributes from a JSON file or URL
- -l, --log-level LOG_LEVEL
- Set the logging level
- -s, --solo
- chef-solo shef session
- -a, --standalone
- standalone shef session
- -v, --version
- Show chef version
- -h, --help
- Show command options
When no --config option is specified, shef attempts to load a default configuration file:
- •
- If a named configuration is given, shef will load ~/.chef/named configuration/shef.rb
- •
- If no named configuration is given shef will load ~/.chef/shef.rb if it exists
- •
- Shef falls back to loading /etc/chef/client.rb or /etc/chef/solo.rb if -z or -s options are given and no shef.rb can be found.
- •
- The --config option takes precedence over implicit configuration paths.
DESCRIPTION
shef is an irb(1) (interactive ruby) session customized for Chef. shef serves two primary functions: it provides a means to interact with a Chef Server interactively using a convenient DSL; it allows you to define and run Chef recipes interactively.SYNTAX
Shef uses irb's subsession feature to provide multiple modes of interaction. In addition to the primary mode which is entered on start, recipe and attributes modes are available.PRIMARY MODE
The following commands are available in the primary session:- help
- Prints a list of available commands
- version
- Prints the Chef version
- recipe
- Switches to recipe mode
- attributes
- Switches to attributes mode
- run_chef
- Initiates a chef run
- reset
- reinitializes shef
- echo :on|:off
- Turns irb's echo function on or off. Echo is on by default.
- tracing :on|:off
- Turns irb's function tracing feature on or off. Tracing is extremely verbose and expected to be of interest primarily to developers.
- node
- Returns the node object for the current host. See knife-node(1) for more information about nodes.
- ohai
- Prints the attributes of node
In addition to these commands, shef provides a DSL for accessing data on the Chef Server. When working with remote data in shef, you chain method calls in the form object type.operation, where object type is in plural form. The following object types are available:
- •
- nodes
- •
- roles
- •
- data_bags
- •
- clients
- •
- cookbooks
For each object type the following operations are available:
- object type.all(&block)
- Loads all items from the server. If the optional code block is given, each item will be passed to the block and the results returned, similar to ruby's Enumerable#map method.
- object type.show(object name)
- Aliased as object type.load
- Loads the singular item identified by object name.
- object type.search(query, &block)
- Aliased as object type.find
- Runs a search against the server and returns the matching items. If the optional code block is given each item will be passed to the block and the results returned.
- The query may be a Solr/Lucene format query given as a String, or a Hash of conditions. If a Hash is given, the options will be ANDed together. To join conditions with OR, use negative queries, or any advanced search syntax, you must provide give the query in String form.
- object type.transform(:all|query, &block)
- Aliased as object type.bulk_edit
- Bulk edit objects by processing them with the (required) code block. You can edit all objects of the given type by passing the Symbol :all as the argument, or only a subset by passing a query as the argument. The query is evaluated in the same way as with search.
- The return value of the code block is used to alter the behavior of transform. If the value returned from the block is nil or false, the object will not be saved. Otherwise, the object is saved after being passed to the block. This behavior can be exploited to create a dry run to test a data transformation.
RECIPE MODE
Recipe mode implements Chef's recipe DSL. Exhaustively documenting this DSL is outside the scope of this document. See the following pages in the Chef documentation for more information:Once you have defined resources in the recipe, you can trigger a convergence run via run_chef
EXAMPLES
- •
- A "Hello World" interactive recipe
-
chef > recipe
chef:recipe > echo :off
chef:recipe > file "/tmp/hello_world"
chef:recipe > run_chef
[Sat, 09 Apr 2011 08:56:56 -0700] INFO: Processing file[/tmp/hello_world] action create ((irb#1) line 2)
[Sat, 09 Apr 2011 08:56:56 -0700] INFO: file[/tmp/hello_world] created file /tmp/hello_world
chef:recipe > pp ls '/tmp'
[".",
"..",
"hello_world"] - •
- Search for nodes by role, and print their IP addresses
-
chef > nodes.find(:roles => 'monitoring-server') {|n| n[:ipaddress] }
=> ["10.254.199.5"] - •
- Remove the role obsolete from every node in the system
-
chef > nodes.transform(:all) {|n| n.run_list.delete('role[obsolete]') }
=> [node[chef098b2.opschef.com], node[ree-woot], node[graphite-dev], node[fluke.localdomain], node[ghost.local], node[kallistec]]
BUGS
The name shef is clever in print but is confusing when spoken aloud. Pronouncing shef as chef console is an imperfect workaround.shef often does not perfectly replicate the context in which chef-client(8) configures a host, which may lead to discrepancies in observed behavior.
shef has to duplicate much code from chef-client's internal libraries and may become out of sync with the behavior of those libraries.