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


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.


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

The Salt Loader

Salt's loader system is responsible for reading Special Module Contents and providing the context for the special Dunder Dictionaries. When modules developed for Salt's loader are imported directly, the dunder attributes won't be populated. You can use the Loader Context to work around this.

Loader Context

Given the following.


def utils_func_getter(name):
    return __utils__[name]

You would not be able import coolmod and run utils_func_getter because __utils__ would not be defined. You must run coolmod.utils_func_getter in the context of a loader.

import coolmod
import salt.loader

opts = {}
utils = salt.loader.utils(opts)
with salt.loader.context(utils):
    func = coolmod.utils_func_getter("")

Special Module Contents

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





Called before __virtual__()


grains and proxy modules

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




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__


Changed in version 3006.0: The __opts__ dictionary can now be accessed via context`.

Defined in: All modules

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


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


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


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__[""]("fdisk -l")


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


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.


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.


Filled in for: Pillar

The __ext_pillar__ dictionary contains the external pillar modules.


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.

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


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.


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


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


Defined in: Engine, Roster, Thorium


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


Defined in: Executor


Defined in: Proxy


Defined in: Thorium


Defined in: Renderers, State


Defined in: State


Defined in: SDB


Changed in version 3006.5.

The __file_client__ dunder was added to states and execution modules. This enables the use of a file client without haveing to instantiate one in the module.