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/target.py

660 lines
21 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.
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
"""
Helper functions intended to be executed on the target. These are entrypoints
for file transfer, module execution and sundry bits like changing file modes.
"""
from __future__ import absolute_import
from __future__ import unicode_literals
import errno
import functools
import grp
import json
import logging
import operator
import os
import pwd
import re
import resource
import signal
import stat
7 years ago
import subprocess
import sys
import tempfile
import traceback
import types
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
import mitogen.core
import mitogen.fork
import mitogen.parent
import mitogen.service
# Ansible since PR #41749 inserts "import __main__" into
# ansible.module_utils.basic. Mitogen's importer will refuse such an import, so
# we must setup a fake "__main__" before that module is ever imported. The
# str() is to cast Unicode to bytes on Python 2.6.
if not sys.modules.get(str('__main__')):
sys.modules[str('__main__')] = types.ModuleType(str('__main__'))
import ansible.module_utils.json_utils
import ansible_mitogen.runner
LOG = logging.getLogger(__name__)
MAKE_TEMP_FAILED_MSG = (
"Unable to find a useable temporary directory. This likely means no\n"
"system-supplied TMP directory can be written to, or all directories\n"
"were mounted on 'noexec' filesystems.\n"
"\n"
"The following paths were tried:\n"
" %(namelist)s\n"
"\n"
"Please check '-vvv' output for a log of individual path errors."
)
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
#: Initialized to an econtext.parent.Context pointing at a pristine fork of
#: the target Python interpreter before it executes any code or imports.
_fork_parent = None
#: Set by :func:`init_child` to the name of a writeable and executable
#: temporary directory accessible by the active user account.
good_temp_dir = None
# issue #362: subprocess.Popen(close_fds=True) aka. AnsibleModule.run_command()
# loops the entire SC_OPEN_MAX space. CentOS>5 ships with 1,048,576 FDs by
# default, resulting in huge (>500ms) runtime waste running many commands.
# Therefore if we are a child, cap the range to something reasonable.
rlimit = resource.getrlimit(resource.RLIMIT_NOFILE)
if (rlimit[0] > 512 or rlimit[1] > 512) and not mitogen.is_master:
resource.setrlimit(resource.RLIMIT_NOFILE, (512, 512))
subprocess.MAXFD = 512 # Python <3.x
del rlimit
def get_small_file(context, path):
"""
Basic in-memory caching module fetcher. This generates an one roundtrip for
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
every previously unseen file, so it is only a temporary solution.
:param context:
Context we should direct FileService requests to. For now (and probably
forever) this is just the top-level Mitogen connection manager process.
:param path:
Path to fetch from FileService, must previously have been registered by
a privileged context using the `register` command.
:returns:
Bytestring file data.
"""
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
pool = mitogen.service.get_or_create_pool(router=context.router)
service = pool.get_service('mitogen.service.PushFileService')
return service.get(path)
def transfer_file(context, in_path, out_path, sync=False, set_owner=False):
"""
Streamily download a file from the connection multiplexer process in the
controller.
:param mitogen.core.Context context:
Reference to the context hosting the FileService that will be used to
fetch the file.
:param bytes in_path:
FileService registered name of the input file.
:param bytes out_path:
Name of the output path on the local disk.
:param bool sync:
If :data:`True`, ensure the file content and metadat are fully on disk
before renaming the temporary file over the existing file. This should
ensure in the case of system crash, either the entire old or new file
are visible post-reboot.
:param bool set_owner:
If :data:`True`, look up the metadata username and group on the local
system and file the file owner using :func:`os.fchmod`.
"""
out_path = os.path.abspath(out_path)
fd, tmp_path = tempfile.mkstemp(suffix='.tmp',
prefix='.ansible_mitogen_transfer-',
dir=os.path.dirname(out_path))
fp = os.fdopen(fd, 'wb', mitogen.core.CHUNK_SIZE)
LOG.debug('transfer_file(%r) temporary file: %s', out_path, tmp_path)
try:
try:
ok, metadata = mitogen.service.FileService.get(
context=context,
path=in_path,
out_fp=fp,
)
if not ok:
raise IOError('transfer of %r was interrupted.' % (in_path,))
os.fchmod(fp.fileno(), metadata['mode'])
if set_owner:
set_fd_owner(fp.fileno(), metadata['owner'], metadata['group'])
finally:
fp.close()
if sync:
os.fsync(fp.fileno())
os.rename(tmp_path, out_path)
except BaseException:
os.unlink(tmp_path)
raise
os.utime(out_path, (metadata['atime'], metadata['mtime']))
def prune_tree(path):
"""
Like shutil.rmtree(), but log errors rather than discard them, and do not
waste multiple os.stat() calls discovering whether the object can be
deleted, just try deleting it instead.
"""
try:
os.unlink(path)
return
except OSError as e:
if not (os.path.isdir(path) and
e.args[0] in (errno.EPERM, errno.EISDIR)):
LOG.error('prune_tree(%r): %s', path, e)
return
try:
# Ensure write access for readonly directories. Ignore error in case
# path is on a weird filesystem (e.g. vfat).
os.chmod(path, int('0700', 8))
except OSError as e:
LOG.warning('prune_tree(%r): %s', path, e)
try:
for name in os.listdir(path):
if name not in ('.', '..'):
prune_tree(os.path.join(path, name))
os.rmdir(path)
except OSError as e:
LOG.error('prune_tree(%r): %s', path, e)
def _on_broker_shutdown():
"""
Respond to broker shutdown (graceful termination by parent, or loss of
connection to parent) by deleting our sole temporary directory.
"""
prune_tree(temp_dir)
def find_good_temp_dir(candidate_temp_dirs):
"""
Given a list of candidate temp directories extracted from ``ansible.cfg``,
combine it with the Python-builtin list of candidate directories used by
:mod:`tempfile`, then iteratively try each until one is found that is both
writeable and executable.
:param list candidate_temp_dirs:
List of candidate $variable-expanded and tilde-expanded directory paths
that may be usable as a temporary directory.
"""
paths = [os.path.expandvars(os.path.expanduser(p))
for p in candidate_temp_dirs]
paths.extend(tempfile._candidate_tempdir_list())
for path in paths:
try:
tmp = tempfile.NamedTemporaryFile(
prefix='ansible_mitogen_find_good_temp_dir',
dir=path,
)
except (OSError, IOError) as e:
LOG.debug('temp dir %r unusable: %s', path, e)
continue
try:
try:
os.chmod(tmp.name, int('0700', 8))
except OSError as e:
LOG.debug('temp dir %r unusable: %s: chmod failed: %s',
path, e)
continue
try:
# access(.., X_OK) is sufficient to detect noexec.
if not os.access(tmp.name, os.X_OK):
raise OSError('filesystem appears to be mounted noexec')
except OSError as e:
LOG.debug('temp dir %r unusable: %s: %s', path, e)
continue
LOG.debug('Selected temp directory: %r (from %r)', path, paths)
return path
finally:
tmp.close()
raise IOError(MAKE_TEMP_FAILED_MSG % {
'paths': '\n '.join(paths),
})
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
@mitogen.core.takes_econtext
def init_child(econtext, log_level, candidate_temp_dirs):
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
"""
Called by ContextService immediately after connection; arranges for the
(presently) spotless Python interpreter to be forked, where the newly
forked interpreter becomes the parent of any newly forked future
interpreters.
This is necessary to prevent modules that are executed in-process from
polluting the global interpreter state in a way that effects explicitly
isolated modules.
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
:param int log_level:
Logging package level active in the master.
:param list[str] candidate_temp_dirs:
List of $variable-expanded and tilde-expanded directory names to add to
candidate list of temporary directories.
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
:returns:
Dict like::
{
'fork_context': mitogen.core.Context.
'home_dir': str.
}
Where `fork_context` refers to the newly forked 'fork parent' context
the controller will use to start forked jobs, and `home_dir` is the
home directory for the active user account.
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
"""
# Copying the master's log level causes log messages to be filtered before
# they reach LogForwarder, thus reducing an influx of tiny messges waking
# the connection multiplexer process in the master.
LOG.setLevel(log_level)
logging.getLogger('ansible_mitogen').setLevel(log_level)
global _fork_parent
mitogen.parent.upgrade_router(econtext)
_fork_parent = econtext.router.fork()
global good_temp_dir
good_temp_dir = find_good_temp_dir(candidate_temp_dirs)
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 {
'fork_context': _fork_parent,
'home_dir': mitogen.core.to_text(os.path.expanduser('~')),
'good_temp_dir': good_temp_dir,
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
}
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
@mitogen.core.takes_econtext
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 create_fork_child(econtext):
"""
For helper functions executed in the fork parent context, arrange for
the context's router to be upgraded as necessary and for a new child to be
prepared.
"""
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
mitogen.parent.upgrade_router(econtext)
context = econtext.router.fork()
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
LOG.debug('create_fork_child() -> %r', context)
return context
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 run_module(kwargs):
"""
Set up the process environment in preparation for running an Ansible
module. This monkey-patches the Ansible libraries in various places to
prevent it from trying to kill the process on completion, and to prevent it
from reading sys.stdin.
"""
runner_name = kwargs.pop('runner_name')
klass = getattr(ansible_mitogen.runner, runner_name)
impl = klass(**kwargs)
return impl.run()
def _get_async_dir():
return os.path.expanduser(
os.environ.get('ANSIBLE_ASYNC_DIR', '~/.ansible_async')
)
class AsyncRunner(object):
def __init__(self, job_id, timeout_secs, econtext, kwargs):
self.job_id = job_id
self.timeout_secs = timeout_secs
self.econtext = econtext
self.kwargs = kwargs
self._timed_out = False
self._init_path()
def _init_path(self):
async_dir = _get_async_dir()
if not os.path.exists(async_dir):
os.makedirs(async_dir)
self.path = os.path.join(async_dir, self.job_id)
def _update(self, dct):
"""
Update an async job status file.
"""
LOG.info('%r._update(%r, %r)', self, self.job_id, dct)
dct.setdefault('ansible_job_id', self.job_id)
dct.setdefault('data', '')
with open(self.path + '.tmp', 'w') as fp:
fp.write(json.dumps(dct))
os.rename(self.path + '.tmp', self.path)
def _on_sigalrm(self, signum, frame):
"""
Respond to SIGALRM (job timeout) by updating the job file and killing
the process.
"""
msg = "Job reached maximum time limit of %d seconds." % (
self.timeout_secs,
)
self._update({
"failed": 1,
"finished": 1,
"msg": msg,
})
self._timed_out = True
self.econtext.broker.shutdown()
def _install_alarm(self):
signal.signal(signal.SIGALRM, self._on_sigalrm)
signal.alarm(self.timeout_secs)
def _run_module(self):
kwargs = dict(self.kwargs, **{
'detach': True,
'econtext': self.econtext,
'emulate_tty': False,
})
return run_module(kwargs)
def _parse_result(self, dct):
filtered, warnings = (
ansible.module_utils.json_utils.
_filter_non_json_lines(dct['stdout'])
)
result = json.loads(filtered)
result.setdefault('warnings', []).extend(warnings)
result['stderr'] = dct['stderr']
self._update(result)
def _run(self):
"""
1. Immediately updates the status file to mark the job as started.
2. Installs a timer/signal handler to implement the time limit.
3. Runs as with run_module(), writing the result to the status file.
:param dict kwargs:
Runner keyword arguments.
:param str job_id:
String job ID.
:param int timeout_secs:
If >0, limit the task's maximum run time.
"""
self._update({
'started': 1,
'finished': 0,
'pid': os.getpid()
})
if self.timeout_secs > 0:
self._install_alarm()
dct = self._run_module()
if not self._timed_out:
# After SIGALRM fires, there is a window between broker responding
# to shutdown() by killing the process, and work continuing on the
# main thread. If main thread was asleep in at least
# basic.py/select.select(), an EINTR will be raised. We want to
# discard that exception.
try:
self._parse_result(dct)
except Exception:
self._update({
"failed": 1,
"msg": traceback.format_exc(),
"data": dct['stdout'], # temporary notice only
"stderr": dct['stderr']
})
def run(self):
try:
try:
self._run()
except Exception:
self._update({
"failed": 1,
"msg": traceback.format_exc(),
})
finally:
self.econtext.broker.shutdown()
@mitogen.core.takes_econtext
def run_module_async(kwargs, job_id, timeout_secs, econtext):
"""
Execute a module with its run status and result written to a file,
terminating on the process on completion. This function must run in a child
forked using :func:`create_fork_child`.
"""
arunner = AsyncRunner(job_id, timeout_secs, econtext, kwargs)
arunner.run()
def get_user_shell():
"""
For commands executed directly via an SSH command-line, SSH looks up the
user's shell via getpwuid() and only defaults to /bin/sh if that field is
missing or empty.
"""
try:
pw_shell = pwd.getpwuid(os.geteuid()).pw_shell
except KeyError:
pw_shell = None
return pw_shell or '/bin/sh'
def exec_args(args, in_data='', chdir=None, shell=None, emulate_tty=False):
"""
Run a command in a subprocess, emulating the argument handling behaviour of
SSH.
:param list[str]:
Argument vector.
:param bytes in_data:
Optional standard input for the command.
:param bool emulate_tty:
If :data:`True`, arrange for stdout and stderr to be merged into the
stdout pipe and for LF to be translated into CRLF, emulating the
behaviour of a TTY.
:return:
(return code, stdout bytes, stderr bytes)
"""
LOG.debug('exec_args(%r, ..., chdir=%r)', args, chdir)
assert isinstance(args, list)
if emulate_tty:
stderr = subprocess.STDOUT
else:
stderr = subprocess.PIPE
proc = subprocess.Popen(
args=args,
7 years ago
stdout=subprocess.PIPE,
stderr=stderr,
7 years ago
stdin=subprocess.PIPE,
cwd=chdir,
)
7 years ago
stdout, stderr = proc.communicate(in_data)
if emulate_tty:
stdout = stdout.replace(b'\n', b'\r\n')
return proc.returncode, stdout, stderr or ''
7 years ago
def exec_command(cmd, in_data='', chdir=None, shell=None, emulate_tty=False):
"""
Run a command in a subprocess, emulating the argument handling behaviour of
SSH.
:param bytes cmd:
String command line, passed to user's shell.
:param bytes in_data:
Optional standard input for the command.
:return:
(return code, stdout bytes, stderr bytes)
"""
assert isinstance(cmd, mitogen.core.UnicodeType)
return exec_args(
args=[get_user_shell(), '-c', cmd],
in_data=in_data,
chdir=chdir,
shell=shell,
emulate_tty=emulate_tty,
)
7 years ago
def read_path(path):
"""
Fetch the contents of a filesystem `path` as bytes.
"""
return open(path, 'rb').read()
7 years ago
def set_fd_owner(fd, owner, group=None):
if owner:
uid = pwd.getpwnam(owner).pw_uid
else:
uid = os.geteuid()
if group:
gid = grp.getgrnam(group).gr_gid
else:
gid = os.getegid()
os.fchown(fd, (uid, gid))
def write_path(path, s, owner=None, group=None, mode=None,
utimes=None, sync=False):
"""
Writes bytes `s` to a filesystem `path`.
"""
path = os.path.abspath(path)
fd, tmp_path = tempfile.mkstemp(suffix='.tmp',
prefix='.ansible_mitogen_transfer-',
dir=os.path.dirname(path))
fp = os.fdopen(fd, 'wb', mitogen.core.CHUNK_SIZE)
LOG.debug('write_path(path=%r) temporary file: %s', path, tmp_path)
try:
try:
if mode:
os.fchmod(fp.fileno(), mode)
if owner or group:
set_fd_owner(fp.fileno(), owner, group)
fp.write(s)
finally:
fp.close()
if sync:
os.fsync(fp.fileno())
os.rename(tmp_path, path)
except BaseException:
os.unlink(tmp_path)
raise
if utimes:
os.utime(path, utimes)
CHMOD_CLAUSE_PAT = re.compile(r'([uoga]*)([+\-=])([ugo]|[rwx]*)')
CHMOD_MASKS = {
'u': stat.S_IRWXU,
'g': stat.S_IRWXG,
'o': stat.S_IRWXO,
'a': (stat.S_IRWXU | stat.S_IRWXG | stat.S_IRWXO),
}
CHMOD_BITS = {
'u': {'r': stat.S_IRUSR, 'w': stat.S_IWUSR, 'x': stat.S_IXUSR},
'g': {'r': stat.S_IRGRP, 'w': stat.S_IWGRP, 'x': stat.S_IXGRP},
'o': {'r': stat.S_IROTH, 'w': stat.S_IWOTH, 'x': stat.S_IXOTH},
'a': {
'r': (stat.S_IRUSR | stat.S_IRGRP | stat.S_IROTH),
'w': (stat.S_IWUSR | stat.S_IWGRP | stat.S_IWOTH),
'x': (stat.S_IXUSR | stat.S_IXGRP | stat.S_IXOTH)
}
}
def apply_mode_spec(spec, mode):
"""
Given a symbolic file mode change specification in the style of chmod(1)
`spec`, apply changes in the specification to the numeric file mode `mode`.
"""
for clause in spec.split(','):
match = CHMOD_CLAUSE_PAT.match(clause)
who, op, perms = match.groups()
for ch in who or 'a':
mask = CHMOD_MASKS[ch]
bits = CHMOD_BITS[ch]
cur_perm_bits = mode & mask
new_perm_bits = functools.reduce(operator.or_, (bits[p] for p in perms), 0)
mode &= ~mask
if op == '=':
mode |= new_perm_bits
elif op == '+':
mode |= new_perm_bits | cur_perm_bits
else:
mode |= cur_perm_bits & ~new_perm_bits
return mode
def set_file_mode(path, spec):
"""
Update the permissions of a file using the same syntax as chmod(1).
"""
mode = os.stat(path).st_mode
if spec.isdigit():
new_mode = int(spec, 8)
else:
new_mode = apply_mode_spec(spec, mode)
os.chmod(path, new_mode)