.. _writing-execution-modules: ========================= Writing Execution Modules ========================= Salt execution modules are the functions called by the :command:`salt` command. Modules Are Easy to Write! ========================== Writing Salt execution modules is straightforward. A Salt execution module is a Python or `Cython`_ module placed in a directory called ``_modules/`` at the root of the Salt fileserver. When using the default fileserver backend (i.e. :py:mod:`roots `), unless environments are otherwise defined in the :conf_master:`file_roots` config option, the ``_modules/`` directory would be located in ``/srv/salt/_modules`` on most systems. Modules placed in ``_modules/`` will be synced to the minions when any of the following Salt functions are called: * :mod:`state.highstate ` (or :mod:`state.apply ` with no state argument) * :mod:`saltutil.sync_modules ` * :mod:`saltutil.sync_all ` Modules placed in ``_modules/`` will be synced to masters when any of the following Salt runners are called: * :mod:`saltutil.sync_modules ` * :mod:`saltutil.sync_all ` Note that a module's default name is its filename (i.e. ``foo.py`` becomes module ``foo``), but that its name can be overridden by using a :ref:`__virtual__ function `. If a Salt module has errors and cannot be imported, the Salt minion will continue to load without issue and the module with errors will simply be omitted. If adding a Cython module the file must be named ``.pyx`` so that the loader knows that the module needs to be imported as a Cython module. The compilation of the Cython module is automatic and happens when the minion starts, so only the ``*.pyx`` file is required. .. _`Cython`: https://cython.org/ Zip Archives as Modules ======================= Python 2.3 and higher allows developers to directly import zip archives containing Python code. By setting :conf_minion:`enable_zip_modules` to ``True`` in the minion config, the Salt loader will be able to import ``.zip`` files in this fashion. This allows Salt module developers to package dependencies with their modules for ease of deployment, isolation, etc. For a user, Zip Archive modules behave just like other modules. When executing a function from a module provided as the file ``my_module.zip``, a user would call a function within that module as ``my_module.``. Creating a Zip Archive Module ----------------------------- A Zip Archive module is structured similarly to a simple `Python package`_. The ``.zip`` file contains a single directory with the same name as the module. The module code traditionally in ``.py`` goes in ``/__init__.py``. The dependency packages are subdirectories of ``/``. Here is an example directory structure for the ``lumberjack`` module, which has two library dependencies (``sleep`` and ``work``) to be included. .. code-block:: bash modules $ ls -R lumberjack __init__.py sleep work lumberjack/sleep: __init__.py lumberjack/work: __init__.py The contents of ``lumberjack/__init__.py`` show how to import and use these included libraries. .. code-block:: python # Libraries included in lumberjack.zip from lumberjack import sleep, work def is_ok(person): """ Checks whether a person is really a lumberjack """ return sleep.all_night(person) and work.all_day(person) Then, create the zip: .. code-block:: console modules $ zip -r lumberjack lumberjack adding: lumberjack/ (stored 0%) adding: lumberjack/__init__.py (deflated 39%) adding: lumberjack/sleep/ (stored 0%) adding: lumberjack/sleep/__init__.py (deflated 7%) adding: lumberjack/work/ (stored 0%) adding: lumberjack/work/__init__.py (deflated 7%) modules $ unzip -l lumberjack.zip Archive: lumberjack.zip Length Date Time Name -------- ---- ---- ---- 0 08-21-15 20:08 lumberjack/ 348 08-21-15 20:08 lumberjack/__init__.py 0 08-21-15 19:53 lumberjack/sleep/ 83 08-21-15 19:53 lumberjack/sleep/__init__.py 0 08-21-15 19:53 lumberjack/work/ 81 08-21-15 19:21 lumberjack/work/__init__.py -------- ------- 512 6 files Once placed in :conf_master:`file_roots`, Salt users can distribute and use ``lumberjack.zip`` like any other module. .. code-block:: bash $ sudo salt minion1 saltutil.sync_modules minion1: - modules.lumberjack $ sudo salt minion1 lumberjack.is_ok 'Michael Palin' minion1: True .. _`Python package`: https://docs.python.org/2/tutorial/modules.html#packages .. _cross-calling-execution-modules: Cross Calling Execution Modules =============================== All of the Salt execution modules are available to each other and modules can call functions available in other execution modules. The variable ``__salt__`` is packed into the modules after they are loaded into the Salt minion. The ``__salt__`` variable is a :ref:`Python dictionary ` containing all of the Salt functions. Dictionary keys are strings representing the names of the modules and the values are the functions themselves. Salt modules can be cross-called by accessing the value in the ``__salt__`` dict: .. code-block:: python def foo(bar): return __salt__["cmd.run"](bar) This code will call the `run` function in the :mod:`cmd ` module and pass the argument ``bar`` to it. Calling Execution Modules on the Salt Master ============================================ .. versionadded:: 2016.11.0 Execution modules can now also be called via the :command:`salt-run` command using the :ref:`salt runner `. Preloaded Execution Module Data =============================== When interacting with execution modules often it is nice to be able to read information dynamically about the minion or to load in configuration parameters for a module. Salt allows for different types of data to be loaded into the modules by the minion. Grains Data ----------- The values detected by the Salt Grains on the minion are available in a :ref:`Python dictionary ` named ``__grains__`` and can be accessed from within callable objects in the Python modules. To see the contents of the grains dictionary for a given system in your deployment run the :func:`grains.items` function: .. code-block:: bash salt 'hostname' grains.items --output=pprint Any value in a grains dictionary can be accessed as any other Python dictionary. For example, the grain representing the minion ID is stored in the ``id`` key and from an execution module, the value would be stored in ``__grains__['id']``. Module Configuration -------------------- Since parameters for configuring a module may be desired, Salt allows for configuration information from the minion configuration file to be passed to execution modules. Since the minion configuration file is a YAML document, arbitrary configuration data can be passed in the minion config that is read by the modules. It is therefore **strongly** recommended that the values passed in the configuration file match the module name. A value intended for the ``test`` execution module should be named ``test.``. The test execution module contains usage of the module configuration and the default configuration file for the minion contains the information and format used to pass data to the modules. :mod:`salt.modules.test`, :file:`conf/minion`. .. _module_init: ``__init__`` Function --------------------- If you want your module to have different execution modes based on minion configuration, you can use the ``__init__(opts)`` function to perform initial module setup. The parameter ``opts`` is the complete minion configuration, as also available in the ``__opts__`` dict. .. code-block:: python """ Cheese module initialization example """ def __init__(opts): """ Allow foreign imports if configured to do so """ if opts.get("cheese.allow_foreign", False): _enable_foreign_products() Strings and Unicode =================== An execution module author should always assume that strings fed to the module have already decoded from strings into Unicode. In Python 2, these will be of type 'Unicode' and in Python 3 they will be of type ``str``. Calling from a state to other Salt sub-systems, should pass Unicode (or bytes if passing binary data). In the rare event that a state needs to write directly to disk, Unicode should be encoded to a string immediately before writing to disk. An author may use ``__salt_system_encoding__`` to learn what the encoding type of the system is. For example, `'my_string'.encode(__salt_system_encoding__')`. Outputter Configuration ======================= Since execution module functions can return different data, and the way the data is printed can greatly change the presentation, Salt allows for a specific outputter to be set on a function-by-function basis. This is done be declaring an ``__outputter__`` dictionary in the global scope of the module. The ``__outputter__`` dictionary contains a mapping of function names to Salt :ref:`outputters `. .. code-block:: python __outputter__ = {"run": "txt"} This will ensure that the ``txt`` outputter is used to display output from the ``run`` function. .. _virtual-modules: Virtual Modules =============== Virtual modules let you override the name of a module in order to use the same name to refer to one of several similar modules. The specific module that is loaded for a virtual name is selected based on the current platform or environment. For example, packages are managed across platforms using the ``pkg`` module. ``pkg`` is a virtual module name that is an alias for the specific package manager module that is loaded on a specific system (for example, :mod:`yumpkg ` on RHEL/CentOS systems , and :mod:`aptpkg ` on Ubuntu). Virtual module names are set using the ``__virtual__`` function and the :ref:`virtual name `. ``__virtual__`` Function ======================== The ``__virtual__`` function returns either a :ref:`string `, :py:data:`True`, :py:data:`False`, or :py:data:`False` with an :ref:`error string `. If a string is returned then the module is loaded using the name of the string as the virtual name. If ``True`` is returned the module is loaded using the current module name. If ``False`` is returned the module is not loaded. ``False`` lets the module perform system checks and prevent loading if dependencies are not met. Since ``__virtual__`` is called before the module is loaded, ``__salt__`` will be unreliable as not all modules will be available at this point in time. The ``__pillar__`` and ``__grains__`` :ref:`"dunder" dictionaries ` are available however. .. note:: Modules which return a string from ``__virtual__`` that is already used by a module that ships with Salt will _override_ the stock module. .. _modules-error-info: Returning Error Information from ``__virtual__`` ------------------------------------------------ Optionally, Salt plugin modules, such as execution, state, returner, beacon, etc. modules may additionally return a string containing the reason that a module could not be loaded. For example, an execution module called ``cheese`` and a corresponding state module also called ``cheese``, both depending on a utility called ``enzymes`` should have ``__virtual__`` functions that handle the case when the dependency is unavailable. .. code-block:: python """ Cheese execution (or returner/beacon/etc.) module """ try: import enzymes HAS_ENZYMES = True except ImportError: HAS_ENZYMES = False def __virtual__(): """ only load cheese if enzymes are available """ if HAS_ENZYMES: return "cheese" else: return ( False, "The cheese execution module cannot be loaded: enzymes unavailable.", ) def slice(): pass .. code-block:: python """ Cheese state module. Note that this works in state modules because it is guaranteed that execution modules are loaded first """ def __virtual__(): """ only load cheese if enzymes are available """ # predicate loading of the cheese state on the corresponding execution module if "cheese.slice" in __salt__: return "cheese" else: return False, "The cheese state module cannot be loaded: enzymes unavailable." Examples -------- The package manager modules are among the best examples of using the ``__virtual__`` function. A table of all the virtual ``pkg`` modules can be found :ref:`here `. .. _module-provider-override: Overriding Virtual Module Providers ----------------------------------- Salt often uses OS grains (``os``, ``osrelease``, ``os_family``, etc.) to determine which module should be loaded as the virtual module for ``pkg``, ``service``, etc. Sometimes this OS detection is incomplete, with new distros popping up, existing distros changing init systems, etc. The virtual modules likely to be affected by this are in the list below (click each item for more information): - :ref:`pkg ` - :ref:`service ` - :ref:`user ` - :ref:`shadow ` - :ref:`group ` If Salt is using the wrong module for one of these, first of all, please `report it on the issue tracker`__, so that this issue can be resolved for a future release. To make it easier to troubleshoot, please also provide the :py:mod:`grains.items ` output, taking care to redact any sensitive information. Then, while waiting for the SaltStack development team to fix the issue, Salt can be made to use the correct module using the :conf_minion:`providers` option in the minion config file: .. code-block:: yaml providers: service: systemd pkg: aptpkg The above example will force the minion to use the :py:mod:`systemd ` module to provide service management, and the :py:mod:`aptpkg ` module to provide package management. .. __: https://github.com/saltstack/salt/issues/new Logging Restrictions -------------------- As a rule, logging should not be done anywhere in a Salt module before it is loaded. This rule apples to all code that would run before the ``__virtual__()`` function, as well as the code within the ``__virtual__()`` function itself. If logging statements are made before the virtual function determines if the module should be loaded, then those logging statements will be called repeatedly. This clutters up log files unnecessarily. Exceptions may be considered for logging statements made at the ``trace`` level. However, it is better to provide the necessary information by another means. One method is to :ref:`return error information ` in the ``__virtual__()`` function. .. _modules-virtual-name: ``__virtualname__`` =================== ``__virtualname__`` is a variable that is used by the documentation build system to know the virtual name of a module without calling the ``__virtual__`` function. Modules that return a string from the ``__virtual__`` function must also set the ``__virtualname__`` variable. To avoid setting the virtual name string twice, you can implement ``__virtual__`` to return the value set for ``__virtualname__`` using a pattern similar to the following: .. code-block:: python # Define the module's virtual name __virtualname__ = "pkg" def __virtual__(): """ Confine this module to Mac OS with Homebrew. """ if salt.utils.path.which("brew") and __grains__["os"] == "MacOS": return __virtualname__ return False The ``__virtual__()`` function can return a ``True`` or ``False`` boolean, a tuple, or a string. If it returns a ``True`` value, this ``__virtualname__`` module-level attribute can be set as seen in the above example. This is the string that the module should be referred to as. When ``__virtual__()`` returns a tuple, the first item should be a boolean and the second should be a string. This is typically done when the module should not load. The first value of the tuple is ``False`` and the second is the error message to display for why the module did not load. For example: .. code-block:: python def __virtual__(): """ Only load if git exists on the system """ if salt.utils.path.which("git") is None: return (False, "The git execution module cannot be loaded: git unavailable.") else: return True Documentation ============= Salt execution modules are documented. The :func:`sys.doc` function will return the documentation for all available modules: .. code-block:: bash salt '*' sys.doc The ``sys.doc`` function simply prints out the docstrings found in the modules; when writing Salt execution modules, please follow the formatting conventions for docstrings as they appear in the other modules. Adding Documentation to Salt Modules ------------------------------------ It is strongly suggested that all Salt modules have documentation added. To add documentation add a `Python docstring`_ to the function. .. code-block:: python def spam(eggs): """ A function to make some spam with eggs! CLI Example:: salt '*' test.spam eggs """ return eggs Now when the sys.doc call is executed the docstring will be cleanly returned to the calling terminal. .. _`Python docstring`: https://docs.python.org/3/glossary.html#term-docstring Documentation added to execution modules in docstrings will automatically be added to the online web-based documentation. Add Execution Module Metadata ----------------------------- When writing a Python docstring for an execution module, add information about the module using the following field lists: .. code-block:: text :maintainer: Thomas Hatch :maturity: new :depends: python-mysqldb :platform: all The maintainer field is a comma-delimited list of developers who help maintain this module. The maturity field indicates the level of quality and testing for this module. Standard labels will be determined. The depends field is a comma-delimited list of modules that this module depends on. The platform field is a comma-delimited list of platforms that this module is known to run on. Log Output ========== You can call the logger from custom modules to write messages to the minion logs. The following code snippet demonstrates writing log messages: .. code-block:: python import logging log = logging.getLogger(__name__) log.info("Here is Some Information") log.warning("You Should Not Do That") log.error("It Is Busted") Aliasing Functions ================== Sometimes one wishes to use a function name that would shadow a python built-in. A common example would be ``set()``. To support this, append an underscore to the function definition, ``def set_():``, and use the ``__func_alias__`` feature to provide an alias to the function. ``__func_alias__`` is a dictionary where each key is the name of a function in the module, and each value is a string representing the alias for that function. When calling an aliased function from a different execution module, state module, or from the cli, the alias name should be used. .. code-block:: python __func_alias__ = { "set_": "set", "list_": "list", } Private Functions ================= In Salt, Python callable objects contained within an execution module are made available to the Salt minion for use. The only exception to this rule is a callable object with a name starting with an underscore ``_``. Objects Loaded Into the Salt Minion ----------------------------------- .. code-block:: python def foo(bar): return bar Objects NOT Loaded into the Salt Minion --------------------------------------- .. code-block:: python def _foobar(baz): # Preceded with an _ return baz cheese = {} # Not a callable Python object Useful Decorators for Modules ============================= Depends Decorator ----------------- When writing execution modules there are many times where some of the module will work on all hosts but some functions have an external dependency, such as a service that needs to be installed or a binary that needs to be present on the system. Instead of trying to wrap much of the code in large try/except blocks, a decorator can be used. If the dependencies passed to the decorator don't exist, then the salt minion will remove those functions from the module on that host. If a ``fallback_function`` is defined, it will replace the function instead of removing it .. code-block:: python import logging from salt.utils.decorators import depends log = logging.getLogger(__name__) try: import dependency_that_sometimes_exists except ImportError as e: log.trace("Failed to import dependency_that_sometimes_exists: {0}".format(e)) @depends("dependency_that_sometimes_exists") def foo(): """ Function with a dependency on the "dependency_that_sometimes_exists" module, if the "dependency_that_sometimes_exists" is missing this function will not exist """ return True def _fallback(): """ Fallback function for the depends decorator to replace a function with """ return '"dependency_that_sometimes_exists" needs to be installed for this function to exist' @depends("dependency_that_sometimes_exists", fallback_function=_fallback) def foo(): """ Function with a dependency on the "dependency_that_sometimes_exists" module. If the "dependency_that_sometimes_exists" is missing this function will be replaced with "_fallback" """ return True In addition to global dependencies the depends decorator also supports raw booleans. .. code-block:: python from salt.utils.decorators import depends HAS_DEP = False try: import dependency_that_sometimes_exists HAS_DEP = True except ImportError: pass @depends(HAS_DEP) def foo(): return True