Developing New Modules

Interactive Debugging

Sometimes debugging with print() and extra logs sprinkled everywhere is not the best strategy.

IPython is a helpful debug tool that has an interactive python environment which can be embedded in python programs.

First the system will require IPython to be installed.

# Debian
apt-get install ipython

# Arch Linux
pacman -Syu ipython2

# RHEL/CentOS (via EPEL)
yum install python-ipython

Now, in the troubling python module, add the following line at a location where the debugger should be started:

test = "test123"
import IPython

IPython.embed_kernel()

After running a Salt command that hits that line, the following will show up in the log file:

[CRITICAL] To connect another client to this kernel, use:
[IPKernelApp] --existing kernel-31271.json

Now on the system that invoked embed_kernel, run the following command from a shell:

# NOTE: use ipython2 instead of ipython for Arch Linux
ipython console --existing

This provides a console that has access to all the vars and functions, and even supports tab-completion.

print(test)
test123

To exit IPython and continue running Salt, press Ctrl-d to logout.

Special Module Contents

These are things that may be defined by the module to influence various things.

__virtual__

__virtual_aliases__

__virtualname__

__init__

Called before __virtual__()

__proxyenabled__

grains and proxy modules

__proxyenabled__ as a list containing the names of the proxy types that the module supports.

__load__

__func_alias__

__outputter__

Dunder Dictionaries

Salt provides several special "dunder" dictionaries as a convenience for Salt development. These include __opts__, __context__, __salt__, and others. This document will describe each dictionary and detail where they exist and what information and/or functionality they provide.

The following dunder dictionaries are always defined, but may be empty

  • __context__

  • __grains__

  • __pillar__

  • __opts__

__opts__

Defined in: All modules

The __opts__ dictionary contains all of the options passed in the configuration file for the master or minion.

Note

In many places in salt, instead of pulling raw data from the __opts__ dict, configuration data should be pulled from the salt get functions such as config.get, aka - __salt__['config.get']('foo:bar') The get functions also allow for dict traversal via the : delimiter. Consider using get functions whenever using __opts__ or __pillar__ and __grains__ (when using grains for configuration data)

The configuration file data made available in the __opts__ dictionary is the configuration data relative to the running daemon. If the modules are loaded and executed by the master, then the master configuration data is available, if the modules are executed by the minion, then the minion configuration is available. Any additional information passed into the respective configuration files is made available

__salt__

Defined in: Auth, Beacons, Engines, Execution, Executors, Outputters, Pillars, Proxies, Renderers, Returners, Runners, SDB, SSH Wrappers, State, Thorium

__salt__ contains the execution module functions. This allows for all functions to be called as they have been set up by the salt loader.

__salt__["cmd.run"]("fdisk -l")
__salt__["network.ip_addrs"]()

Note

When used in runners or outputters, __salt__ references other runner/outputter modules, and not execution modules.

__grains__

Filled in for: Execution, Pillar, Renderer, Returner, SSH Wrapper, State.

The __grains__ dictionary contains the grains data generated by the minion that is currently being worked with. In execution modules, state modules and returners this is the grains of the minion running the calls, when generating the external pillar the __grains__ is the grains data from the minion that the pillar is being generated for.

While __grains__ is defined for every module, it's only filled in for some.

__pillar__

Filled in for: Execution, Renderer, Returner, SSH Wrapper, State

The __pillar__ dictionary contains the pillar for the respective minion.

While __pillar__ is defined for every module, it's only filled in for some.

__ext_pillar__

Filled in for: Pillar

The __ext_pillar__ dictionary contains the external pillar modules.

__context__

During a state run the __context__ dictionary persists across all states that are run and then is destroyed when the state ends.

When running an execution module __context__ persists across all module executions until the modules are refreshed; such as when saltutil.sync_all or state.apply are executed.

A great place to see how to use __context__ is in the cp.py module in salt/modules/cp.py. The fileclient authenticates with the master when it is instantiated and then is used to copy files to the minion. Rather than create a new fileclient for each file that is to be copied down, one instance of the fileclient is instantiated in the __context__ dictionary and is reused for each file. Here is an example from salt/modules/cp.py:

if not "cp.fileclient" in __context__:
    __context__["cp.fileclient"] = salt.fileclient.get_file_client(__opts__)

Note

Because __context__ may or may not have been destroyed, always be sure to check for the existence of the key in __context__ and generate the key before using it.

__utils__

Defined in: Cloud, Engine, Execution, File Server, Grain, Pillar, Proxy, Roster, Runner, SDB, State

__proxy__

Defined in: Beacon, Engine, Execution, Executor, Proxy, Renderer, Returner, State, Util

__runner__

Defined in: Engine, Roster, Thorium

Note

When used in engines, it should be called __runners__ (plural)

__executors__

Defined in: Executor

__ret__

Defined in: Proxy

__thorium__

Defined in: Thorium

__states__

Defined in: Renderers, State

__serializers__

Defined in: State

__sdb__

Defined in: SDB

Additional Globals

Defined for: Runners, Execution Modules, Wheels

  • __jid__: The job ID

  • __user__: The user

  • __tag__: The jid tag

  • __jid_event__: A salt.utils.event.NamespacedEvent.

NamespacedEvent defines a single method fire_event, that takes data and tag. The Runner docs has examples.