You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
mitogen/ansible_mitogen/planner.py

459 lines
15 KiB
Python

# Copyright 2017, David Wilson
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
#
# 1. Redistributions of source code must retain the above copyright notice,
# this list of conditions and the following disclaimer.
#
# 2. Redistributions in binary form must reproduce the above copyright notice,
# this list of conditions and the following disclaimer in the documentation
# and/or other materials provided with the distribution.
#
# 3. Neither the name of the copyright holder nor the names of its contributors
# may be used to endorse or promote products derived from this software without
# specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
# POSSIBILITY OF SUCH DAMAGE.
"""
Classes to detect each case from [0] and prepare arguments necessary for the
corresponding Runner class within the target, including preloading requisite
files/modules known missing.
[0] "Ansible Module Architecture", developing_program_flow_modules.html
"""
from __future__ import absolute_import
from __future__ import unicode_literals
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
import json
import logging
import os
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
import random
from ansible.executor import module_common
import ansible.errors
import ansible.module_utils
import mitogen.core
import ansible_mitogen.loaders
import ansible_mitogen.parsing
import ansible_mitogen.target
LOG = logging.getLogger(__name__)
NO_METHOD_MSG = 'Mitogen: no invocation method found for: '
NO_INTERPRETER_MSG = 'module (%s) is missing interpreter line'
class Invocation(object):
"""
Collect up a module's execution environment then use it to invoke
target.run_module() or helpers.run_module_async() in the target context.
"""
def __init__(self, action, connection, module_name, module_args,
task_vars, templar, env, wrap_async, timeout_secs):
#: ActionBase instance invoking the module. Required to access some
#: output postprocessing methods that don't belong in ActionBase at
#: all.
self.action = action
#: Ansible connection to use to contact the target. Must be an
#: ansible_mitogen connection.
self.connection = connection
#: Name of the module ('command', 'shell', etc.) to execute.
self.module_name = module_name
#: Final module arguments.
self.module_args = module_args
#: Task variables, needed to extract ansible_*_interpreter.
self.task_vars = task_vars
#: Templar, needed to extract ansible_*_interpreter.
self.templar = templar
#: Final module environment.
self.env = env
#: Boolean, if :py:data:`True`, launch the module asynchronously.
self.wrap_async = wrap_async
#: Integer, if >0, limit the time an asynchronous job may run for.
self.timeout_secs = timeout_secs
#: Initially ``None``, but set by :func:`invoke`. The path on the
#: master to the module's implementation file.
self.module_path = None
#: Initially ``None``, but set by :func:`invoke`. The raw source or
#: binary contents of the module.
self.module_source = None
def __repr__(self):
return 'Invocation(module_name=%s)' % (self.module_name,)
class Planner(object):
"""
A Planner receives a module name and the contents of its implementation
file, indicates whether or not it understands how to run the module, and
exports a method to run the module.
"""
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def __init__(self, invocation):
self._inv = invocation
def detect(self):
"""
Return true if the supplied `invocation` matches the module type
implemented by this planner.
"""
raise NotImplementedError()
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def should_fork(self):
"""
Asynchronous tasks must always be forked.
"""
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
return self._inv.wrap_async
def get_push_files(self):
"""
Return a list of files that should be propagated to the target context
using PushFileService. The default implementation pushes nothing.
"""
return []
def get_module_deps(self):
"""
Return a list of the Python module names imported by the module.
"""
return []
ansible: enable forking when requested and for async jobs. Closes #105. References #155. mitogen/service.py: Refactor services to support individually exposed methods with different security policies for each method. - @mitogen.service.expose() to expose a method and set its policy - @mitogen.service.arg_spec() to validate input. - Require basic service message format to be a tuple of `(method, kwargs)`, where kwargs is always a dict. - Update DeduplicatingService to match the new scheme. ansible_mitogen/connection.py: - Rename 'method' to 'method_name' to disambiguate it from the service.call()'s method= argument. ansible_mitogen/planner.py: - Generate an ID for every job, sync or not, and fetch job results from JobResultService rather than via the initiating function call's return value. - Planner subclasses now get to select whether their Runner should run in a forked process. The base implementation requests this if the 'mitogen_isolation_mode=fork' task variable is present. ansible_mitogen/runner.py: Teach runners to deliver their result via JobResultService executing in their indirect parent mux process. ansible_mitogen/plugins/actions/mitogen_async_status.py: Split the implementation up into methods, and more compatibly emulate Ansible's existing output. ansible_mitogen/process.py: Mux processes now host JobResultService. ansible_mitogen/services.py: Update existing services to the new mitogen.service scheme, and implement JobResultService: * listen() method for synchronous jobs. planner.invoke() registers a Sender with the service prior to invoking the job, then sleeps waiting for the service to write the job result to the corresponding Receiver. * Non-blocking get() method for implementing mitogen_async_status action. * Child-accessible push() method for delivering task results. ansible_mitogen/target.py: New helpers for spawning a virginal subprocess on startup, from which asynchronous and mitogen_task_isolation=fork jobs are forked. Necessary to avoid a task inheriting potentially polluted/monkey-patched parent environment, since remaining jobs continue to run in the original child process. docs/ansible.rst: Add/merge/remove some behaviours/risks. tests/ansible/integration: New tests for forking/async.
7 years ago
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def get_kwargs(self, **kwargs):
"""
If :meth:`detect` returned :data:`True`, plan for the module's
execution, including granting access to or delivering any files to it
that are known to be absent, and finally return a dict::
{
# Name of the class from runners.py that implements the
# target-side execution of this module type.
"runner_name": "...",
# Remaining keys are passed to the constructor of the class
# named by `runner_name`.
}
"""
new = dict((mitogen.core.UnicodeType(k), kwargs[k])
for k in kwargs)
new.setdefault('cwd', self._inv.connection.get_default_cwd())
new.setdefault('extra_env', self._inv.connection.get_default_env())
new.setdefault('emulate_tty', True)
new.setdefault('service_context', self._inv.connection.parent)
return new
def __repr__(self):
return '%s()' % (type(self).__name__,)
class BinaryPlanner(Planner):
"""
Binary modules take their arguments and will return data to Ansible in the
same way as want JSON modules.
"""
runner_name = 'BinaryRunner'
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def detect(self):
return module_common._is_binary(self._inv.module_source)
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def get_push_files(self):
return [self._inv.module_path]
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def get_kwargs(self, **kwargs):
return super(BinaryPlanner, self).get_kwargs(
ansible: enable forking when requested and for async jobs. Closes #105. References #155. mitogen/service.py: Refactor services to support individually exposed methods with different security policies for each method. - @mitogen.service.expose() to expose a method and set its policy - @mitogen.service.arg_spec() to validate input. - Require basic service message format to be a tuple of `(method, kwargs)`, where kwargs is always a dict. - Update DeduplicatingService to match the new scheme. ansible_mitogen/connection.py: - Rename 'method' to 'method_name' to disambiguate it from the service.call()'s method= argument. ansible_mitogen/planner.py: - Generate an ID for every job, sync or not, and fetch job results from JobResultService rather than via the initiating function call's return value. - Planner subclasses now get to select whether their Runner should run in a forked process. The base implementation requests this if the 'mitogen_isolation_mode=fork' task variable is present. ansible_mitogen/runner.py: Teach runners to deliver their result via JobResultService executing in their indirect parent mux process. ansible_mitogen/plugins/actions/mitogen_async_status.py: Split the implementation up into methods, and more compatibly emulate Ansible's existing output. ansible_mitogen/process.py: Mux processes now host JobResultService. ansible_mitogen/services.py: Update existing services to the new mitogen.service scheme, and implement JobResultService: * listen() method for synchronous jobs. planner.invoke() registers a Sender with the service prior to invoking the job, then sleeps waiting for the service to write the job result to the corresponding Receiver. * Non-blocking get() method for implementing mitogen_async_status action. * Child-accessible push() method for delivering task results. ansible_mitogen/target.py: New helpers for spawning a virginal subprocess on startup, from which asynchronous and mitogen_task_isolation=fork jobs are forked. Necessary to avoid a task inheriting potentially polluted/monkey-patched parent environment, since remaining jobs continue to run in the original child process. docs/ansible.rst: Add/merge/remove some behaviours/risks. tests/ansible/integration: New tests for forking/async.
7 years ago
runner_name=self.runner_name,
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
module=self._inv.module_name,
path=self._inv.module_path,
json_args=json.dumps(self._inv.module_args),
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
env=self._inv.env,
ansible: enable forking when requested and for async jobs. Closes #105. References #155. mitogen/service.py: Refactor services to support individually exposed methods with different security policies for each method. - @mitogen.service.expose() to expose a method and set its policy - @mitogen.service.arg_spec() to validate input. - Require basic service message format to be a tuple of `(method, kwargs)`, where kwargs is always a dict. - Update DeduplicatingService to match the new scheme. ansible_mitogen/connection.py: - Rename 'method' to 'method_name' to disambiguate it from the service.call()'s method= argument. ansible_mitogen/planner.py: - Generate an ID for every job, sync or not, and fetch job results from JobResultService rather than via the initiating function call's return value. - Planner subclasses now get to select whether their Runner should run in a forked process. The base implementation requests this if the 'mitogen_isolation_mode=fork' task variable is present. ansible_mitogen/runner.py: Teach runners to deliver their result via JobResultService executing in their indirect parent mux process. ansible_mitogen/plugins/actions/mitogen_async_status.py: Split the implementation up into methods, and more compatibly emulate Ansible's existing output. ansible_mitogen/process.py: Mux processes now host JobResultService. ansible_mitogen/services.py: Update existing services to the new mitogen.service scheme, and implement JobResultService: * listen() method for synchronous jobs. planner.invoke() registers a Sender with the service prior to invoking the job, then sleeps waiting for the service to write the job result to the corresponding Receiver. * Non-blocking get() method for implementing mitogen_async_status action. * Child-accessible push() method for delivering task results. ansible_mitogen/target.py: New helpers for spawning a virginal subprocess on startup, from which asynchronous and mitogen_task_isolation=fork jobs are forked. Necessary to avoid a task inheriting potentially polluted/monkey-patched parent environment, since remaining jobs continue to run in the original child process. docs/ansible.rst: Add/merge/remove some behaviours/risks. tests/ansible/integration: New tests for forking/async.
7 years ago
**kwargs
)
class ScriptPlanner(BinaryPlanner):
"""
Common functionality for script module planners -- handle interpreter
detection and rewrite.
"""
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def _get_interpreter(self):
interpreter, arg = ansible_mitogen.parsing.parse_hashbang(
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
self._inv.module_source
)
if interpreter is None:
raise ansible.errors.AnsibleError(NO_INTERPRETER_MSG % (
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
self._inv.module_name,
))
key = u'ansible_%s_interpreter' % os.path.basename(interpreter).strip()
try:
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
template = self._inv.task_vars[key].strip()
return self._inv.templar.template(template), arg
except KeyError:
return interpreter, arg
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def get_kwargs(self, **kwargs):
interpreter, arg = self._get_interpreter()
return super(ScriptPlanner, self).get_kwargs(
interpreter_arg=arg,
interpreter=interpreter,
ansible: enable forking when requested and for async jobs. Closes #105. References #155. mitogen/service.py: Refactor services to support individually exposed methods with different security policies for each method. - @mitogen.service.expose() to expose a method and set its policy - @mitogen.service.arg_spec() to validate input. - Require basic service message format to be a tuple of `(method, kwargs)`, where kwargs is always a dict. - Update DeduplicatingService to match the new scheme. ansible_mitogen/connection.py: - Rename 'method' to 'method_name' to disambiguate it from the service.call()'s method= argument. ansible_mitogen/planner.py: - Generate an ID for every job, sync or not, and fetch job results from JobResultService rather than via the initiating function call's return value. - Planner subclasses now get to select whether their Runner should run in a forked process. The base implementation requests this if the 'mitogen_isolation_mode=fork' task variable is present. ansible_mitogen/runner.py: Teach runners to deliver their result via JobResultService executing in their indirect parent mux process. ansible_mitogen/plugins/actions/mitogen_async_status.py: Split the implementation up into methods, and more compatibly emulate Ansible's existing output. ansible_mitogen/process.py: Mux processes now host JobResultService. ansible_mitogen/services.py: Update existing services to the new mitogen.service scheme, and implement JobResultService: * listen() method for synchronous jobs. planner.invoke() registers a Sender with the service prior to invoking the job, then sleeps waiting for the service to write the job result to the corresponding Receiver. * Non-blocking get() method for implementing mitogen_async_status action. * Child-accessible push() method for delivering task results. ansible_mitogen/target.py: New helpers for spawning a virginal subprocess on startup, from which asynchronous and mitogen_task_isolation=fork jobs are forked. Necessary to avoid a task inheriting potentially polluted/monkey-patched parent environment, since remaining jobs continue to run in the original child process. docs/ansible.rst: Add/merge/remove some behaviours/risks. tests/ansible/integration: New tests for forking/async.
7 years ago
**kwargs
)
class JsonArgsPlanner(ScriptPlanner):
"""
Script that has its interpreter directive and the task arguments
substituted into its source as a JSON string.
"""
runner_name = 'JsonArgsRunner'
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def detect(self):
return module_common.REPLACER_JSONARGS in self._inv.module_source
class WantJsonPlanner(ScriptPlanner):
"""
If a module has the string WANT_JSON in it anywhere, Ansible treats it as a
non-native module that accepts a filename as its only command line
parameter. The filename is for a temporary file containing a JSON string
containing the module's parameters. The module needs to open the file, read
and parse the parameters, operate on the data, and print its return data as
a JSON encoded dictionary to stdout before exiting.
These types of modules are self-contained entities. As of Ansible 2.1,
Ansible only modifies them to change a shebang line if present.
"""
runner_name = 'WantJsonRunner'
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def detect(self):
return b'WANT_JSON' in self._inv.module_source
class NewStylePlanner(ScriptPlanner):
"""
The Ansiballz framework differs from module replacer in that it uses real
Python imports of things in ansible/module_utils instead of merely
preprocessing the module.
"""
runner_name = 'NewStyleRunner'
marker = b'from ansible.module_utils.'
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def detect(self):
return self.marker in self._inv.module_source
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def _get_interpreter(self):
return None, None
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def get_push_files(self):
return super(NewStylePlanner, self).get_push_files() + [
path
for fullname, path, is_pkg in self.get_module_map()['custom']
]
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def get_module_deps(self):
return self.get_module_map()['builtin']
#: Module names appearing in this set always require forking, usually due
#: to some terminal leakage that cannot be worked around in any sane
#: manner.
ALWAYS_FORK_MODULES = frozenset([
'dnf', # issue #280; py-dnf/hawkey need therapy
])
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def should_fork(self):
"""
In addition to asynchronous tasks, new-style modules should be forked
if:
* the user specifies mitogen_task_isolation=fork, or
* the new-style module has a custom module search path, or
* the module is known to leak like a sieve.
"""
ansible: enable forking when requested and for async jobs. Closes #105. References #155. mitogen/service.py: Refactor services to support individually exposed methods with different security policies for each method. - @mitogen.service.expose() to expose a method and set its policy - @mitogen.service.arg_spec() to validate input. - Require basic service message format to be a tuple of `(method, kwargs)`, where kwargs is always a dict. - Update DeduplicatingService to match the new scheme. ansible_mitogen/connection.py: - Rename 'method' to 'method_name' to disambiguate it from the service.call()'s method= argument. ansible_mitogen/planner.py: - Generate an ID for every job, sync or not, and fetch job results from JobResultService rather than via the initiating function call's return value. - Planner subclasses now get to select whether their Runner should run in a forked process. The base implementation requests this if the 'mitogen_isolation_mode=fork' task variable is present. ansible_mitogen/runner.py: Teach runners to deliver their result via JobResultService executing in their indirect parent mux process. ansible_mitogen/plugins/actions/mitogen_async_status.py: Split the implementation up into methods, and more compatibly emulate Ansible's existing output. ansible_mitogen/process.py: Mux processes now host JobResultService. ansible_mitogen/services.py: Update existing services to the new mitogen.service scheme, and implement JobResultService: * listen() method for synchronous jobs. planner.invoke() registers a Sender with the service prior to invoking the job, then sleeps waiting for the service to write the job result to the corresponding Receiver. * Non-blocking get() method for implementing mitogen_async_status action. * Child-accessible push() method for delivering task results. ansible_mitogen/target.py: New helpers for spawning a virginal subprocess on startup, from which asynchronous and mitogen_task_isolation=fork jobs are forked. Necessary to avoid a task inheriting potentially polluted/monkey-patched parent environment, since remaining jobs continue to run in the original child process. docs/ansible.rst: Add/merge/remove some behaviours/risks. tests/ansible/integration: New tests for forking/async.
7 years ago
return (
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
super(NewStylePlanner, self).should_fork() or
(self._inv.task_vars.get('mitogen_task_isolation') == 'fork') or
(self._inv.module_name in self.ALWAYS_FORK_MODULES) or
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
(len(self.get_module_map()['custom']) > 0)
ansible: enable forking when requested and for async jobs. Closes #105. References #155. mitogen/service.py: Refactor services to support individually exposed methods with different security policies for each method. - @mitogen.service.expose() to expose a method and set its policy - @mitogen.service.arg_spec() to validate input. - Require basic service message format to be a tuple of `(method, kwargs)`, where kwargs is always a dict. - Update DeduplicatingService to match the new scheme. ansible_mitogen/connection.py: - Rename 'method' to 'method_name' to disambiguate it from the service.call()'s method= argument. ansible_mitogen/planner.py: - Generate an ID for every job, sync or not, and fetch job results from JobResultService rather than via the initiating function call's return value. - Planner subclasses now get to select whether their Runner should run in a forked process. The base implementation requests this if the 'mitogen_isolation_mode=fork' task variable is present. ansible_mitogen/runner.py: Teach runners to deliver their result via JobResultService executing in their indirect parent mux process. ansible_mitogen/plugins/actions/mitogen_async_status.py: Split the implementation up into methods, and more compatibly emulate Ansible's existing output. ansible_mitogen/process.py: Mux processes now host JobResultService. ansible_mitogen/services.py: Update existing services to the new mitogen.service scheme, and implement JobResultService: * listen() method for synchronous jobs. planner.invoke() registers a Sender with the service prior to invoking the job, then sleeps waiting for the service to write the job result to the corresponding Receiver. * Non-blocking get() method for implementing mitogen_async_status action. * Child-accessible push() method for delivering task results. ansible_mitogen/target.py: New helpers for spawning a virginal subprocess on startup, from which asynchronous and mitogen_task_isolation=fork jobs are forked. Necessary to avoid a task inheriting potentially polluted/monkey-patched parent environment, since remaining jobs continue to run in the original child process. docs/ansible.rst: Add/merge/remove some behaviours/risks. tests/ansible/integration: New tests for forking/async.
7 years ago
)
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def get_search_path(self):
return tuple(
path
for path in ansible_mitogen.loaders.module_utils_loader._get_paths(
subdirs=False
)
if os.path.isdir(path)
)
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
_module_map = None
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def get_module_map(self):
if self._module_map is None:
self._module_map = self._inv.connection.parent.call_service(
service_name='ansible_mitogen.services.ModuleDepService',
method_name='scan',
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
module_name='ansible_module_%s' % (self._inv.module_name,),
module_path=self._inv.module_path,
search_path=self.get_search_path(),
builtin_path=module_common._MODULE_UTILS_PATH,
context=self._inv.connection.context,
)
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
return self._module_map
def get_kwargs(self):
return super(NewStylePlanner, self).get_kwargs(
module_map=self.get_module_map(),
)
class ReplacerPlanner(NewStylePlanner):
"""
The Module Replacer framework is the original framework implementing
new-style modules. It is essentially a preprocessor (like the C
Preprocessor for those familiar with that programming language). It does
straight substitutions of specific substring patterns in the module file.
There are two types of substitutions.
* Replacements that only happen in the module file. These are public
replacement strings that modules can utilize to get helpful boilerplate
or access to arguments.
"from ansible.module_utils.MOD_LIB_NAME import *" is replaced with the
contents of the ansible/module_utils/MOD_LIB_NAME.py. These should only
be used with new-style Python modules.
"#<<INCLUDE_ANSIBLE_MODULE_COMMON>>" is equivalent to
"from ansible.module_utils.basic import *" and should also only apply to
new-style Python modules.
"# POWERSHELL_COMMON" substitutes the contents of
"ansible/module_utils/powershell.ps1". It should only be used with
new-style Powershell modules.
"""
runner_name = 'ReplacerRunner'
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def detect(self):
return module_common.REPLACER in self._inv.module_source
class OldStylePlanner(ScriptPlanner):
runner_name = 'OldStyleRunner'
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def detect(self):
# Everything else.
return True
_planners = [
BinaryPlanner,
# ReplacerPlanner,
NewStylePlanner,
JsonArgsPlanner,
WantJsonPlanner,
OldStylePlanner,
]
def get_module_data(name):
path = ansible_mitogen.loaders.module_loader.find_plugin(name, '')
with open(path, 'rb') as fp:
source = fp.read()
return mitogen.core.to_text(path), source
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def _propagate_deps(invocation, planner, context):
invocation.connection.parent.call_service(
service_name='mitogen.service.PushFileService',
method_name='propagate_paths_and_modules',
context=context,
paths=planner.get_push_files(),
modules=planner.get_module_deps(),
)
def _invoke_async_task(invocation, planner):
job_id = '%016x' % random.randint(0, 2**64)
context = invocation.connection.create_fork_child()
_propagate_deps(invocation, planner, context)
context.call_no_reply(
ansible_mitogen.target.run_module_async,
job_id=job_id,
timeout_secs=invocation.timeout_secs,
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
kwargs=planner.get_kwargs(),
)
return {
'stdout': json.dumps({
# modules/utilities/logic/async_wrapper.py::_run_module().
'changed': True,
'started': 1,
'finished': 0,
'ansible_job_id': job_id,
})
}
def _invoke_forked_task(invocation, planner):
context = invocation.connection.create_fork_child()
_propagate_deps(invocation, planner, context)
try:
return context.call(
ansible_mitogen.target.run_module,
kwargs=planner.get_kwargs(),
)
finally:
context.shutdown()
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def _get_planner(invocation):
for klass in _planners:
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
planner = klass(invocation)
if planner.detect():
LOG.debug('%r accepted %r (filename %r)', planner,
invocation.module_name, invocation.module_path)
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
return planner
LOG.debug('%r rejected %r', planner, invocation.module_name)
raise ansible.errors.AnsibleError(NO_METHOD_MSG + repr(invocation))
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
def invoke(invocation):
"""
Find a Planner subclass corresnding to `invocation` and use it to invoke
the module.
:param Invocation invocation:
:returns:
Module return dict.
:raises ansible.errors.AnsibleError:
Unrecognized/unsupported module type.
"""
(invocation.module_path,
invocation.module_source) = get_module_data(invocation.module_name)
planner = _get_planner(invocation)
if invocation.wrap_async:
response = _invoke_async_task(invocation, planner)
elif planner.should_fork():
response = _invoke_forked_task(invocation, planner)
else:
_propagate_deps(invocation, planner, invocation.connection.context)
response = invocation.connection.call(
ansible_mitogen.target.run_module,
kwargs=planner.get_kwargs(),
)
return invocation.action._postprocess_response(response)