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/connection.py

643 lines
22 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.
from __future__ import absolute_import
from __future__ import unicode_literals
import logging
import os
import shlex
import stat
import time
7 years ago
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
import jinja2.runtime
import ansible.constants as C
import ansible.errors
7 years ago
import ansible.plugins.connection
import ansible.utils.shlex
import mitogen.unix
import mitogen.utils
7 years ago
import ansible_mitogen.target
import ansible_mitogen.process
import ansible_mitogen.services
7 years ago
LOG = logging.getLogger(__name__)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
def _connect_local(spec):
return {
'method': 'local',
'kwargs': {
'python_path': spec['python_path'],
}
}
def wrap_or_none(klass, value):
if value is not None:
return klass(value)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
def _connect_ssh(spec):
if C.HOST_KEY_CHECKING:
check_host_keys = 'enforce'
else:
check_host_keys = 'ignore'
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
return {
'method': 'ssh',
'kwargs': {
'check_host_keys': check_host_keys,
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
'hostname': spec['remote_addr'],
'username': spec['remote_user'],
'password': wrap_or_none(mitogen.core.Secret, spec['password']),
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
'port': spec['port'],
'python_path': spec['python_path'],
'identity_file': spec['private_key_file'],
'ssh_path': spec['ssh_executable'],
'connect_timeout': spec['ansible_ssh_timeout'],
'ssh_args': spec['ssh_args'],
'ssh_debug_level': spec['mitogen_ssh_debug_level'],
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
}
}
def _connect_docker(spec):
return {
'method': 'docker',
'kwargs': {
'username': spec['remote_user'],
'container': spec['remote_addr'],
'python_path': spec['python_path'],
'connect_timeout': spec['ansible_ssh_timeout'] or spec['timeout'],
}
}
def _connect_jail(spec):
return {
'method': 'jail',
'kwargs': {
'username': spec['remote_user'],
'container': spec['remote_addr'],
'python_path': spec['python_path'],
'connect_timeout': spec['ansible_ssh_timeout'] or spec['timeout'],
}
}
def _connect_lxc(spec):
return {
'method': 'lxc',
'kwargs': {
'container': spec['remote_addr'],
'python_path': spec['python_path'],
'connect_timeout': spec['ansible_ssh_timeout'] or spec['timeout'],
}
}
def _connect_machinectl(spec):
return _connect_setns(dict(spec, mitogen_kind='machinectl'))
def _connect_setns(spec):
return {
'method': 'setns',
'kwargs': {
'container': spec['remote_addr'],
'username': spec['remote_user'],
'python_path': spec['python_path'],
'kind': spec['mitogen_kind'],
'docker_path': spec['mitogen_docker_path'],
'lxc_info_path': spec['mitogen_lxc_info_path'],
'machinectl_path': spec['mitogen_machinectl_path'],
}
}
def _connect_su(spec):
return {
'method': 'su',
'enable_lru': True,
'kwargs': {
'username': spec['become_user'],
'password': wrap_or_none(mitogen.core.Secret, spec['become_pass']),
'python_path': spec['python_path'],
'su_path': spec['become_exe'],
'connect_timeout': spec['timeout'],
}
}
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
def _connect_sudo(spec):
return {
'method': 'sudo',
'enable_lru': True,
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
'kwargs': {
'username': spec['become_user'],
'password': wrap_or_none(mitogen.core.Secret, spec['become_pass']),
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
'python_path': spec['python_path'],
'sudo_path': spec['become_exe'],
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
'connect_timeout': spec['timeout'],
'sudo_args': spec['sudo_args'],
}
}
def _connect_mitogen_su(spec):
# su as a first-class proxied connection, not a become method.
return {
'method': 'su',
'kwargs': {
'username': spec['remote_user'],
'password': wrap_or_none(mitogen.core.Secret, spec['password']),
'python_path': spec['python_path'],
'su_path': spec['become_exe'],
'connect_timeout': spec['timeout'],
}
}
def _connect_mitogen_sudo(spec):
# sudo as a first-class proxied connection, not a become method.
return {
'method': 'sudo',
'kwargs': {
'username': spec['remote_user'],
'password': wrap_or_none(mitogen.core.Secret, spec['password']),
'python_path': spec['python_path'],
'sudo_path': spec['become_exe'],
'connect_timeout': spec['timeout'],
'sudo_args': spec['sudo_args'],
}
}
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
CONNECTION_METHOD = {
'docker': _connect_docker,
'jail': _connect_jail,
'local': _connect_local,
'lxc': _connect_lxc,
'lxd': _connect_lxc,
'machinectl': _connect_machinectl,
'setns': _connect_setns,
'ssh': _connect_ssh,
'su': _connect_su,
'sudo': _connect_sudo,
'mitogen_su': _connect_mitogen_su,
'mitogen_sudo': _connect_mitogen_sudo,
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
}
def config_from_play_context(transport, inventory_name, connection):
"""
Return a dict representing all important connection configuration, allowing
the same functions to work regardless of whether configuration came from
play_context (direct connection) or host vars (mitogen_via=).
"""
return {
'transport': transport,
'inventory_name': inventory_name,
'remote_addr': connection._play_context.remote_addr,
'remote_user': connection._play_context.remote_user,
'become': connection._play_context.become,
'become_method': connection._play_context.become_method,
'become_user': connection._play_context.become_user,
'become_pass': connection._play_context.become_pass,
'password': connection._play_context.password,
'port': connection._play_context.port,
'python_path': connection.python_path,
'private_key_file': connection._play_context.private_key_file,
'ssh_executable': connection._play_context.ssh_executable,
'timeout': connection._play_context.timeout,
'ansible_ssh_timeout': connection.ansible_ssh_timeout,
'ssh_args': [
mitogen.core.to_text(term)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
for s in (
getattr(connection._play_context, 'ssh_args', ''),
getattr(connection._play_context, 'ssh_common_args', ''),
getattr(connection._play_context, 'ssh_extra_args', '')
)
for term in ansible.utils.shlex.shlex_split(s or '')
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
],
'become_exe': connection._play_context.become_exe,
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
'sudo_args': [
mitogen.core.to_text(term)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
for s in (
connection._play_context.sudo_flags,
connection._play_context.become_flags
)
for term in ansible.utils.shlex.shlex_split(s or '')
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
],
'mitogen_via': connection.mitogen_via,
'mitogen_kind': connection.mitogen_kind,
'mitogen_docker_path': connection.mitogen_docker_path,
'mitogen_lxc_info_path': connection.mitogen_lxc_info_path,
'mitogen_machinectl_path': connection.mitogen_machinectl_path,
'mitogen_ssh_debug_level': connection.mitogen_ssh_debug_level,
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
}
def config_from_hostvars(transport, inventory_name, connection,
hostvars, become_user):
"""
Override config_from_play_context() to take equivalent information from
host vars.
"""
config = config_from_play_context(transport, inventory_name, connection)
hostvars = dict(hostvars)
return dict(config, **{
'remote_addr': hostvars.get('ansible_hostname', inventory_name),
'become': bool(become_user),
'become_user': become_user,
'become_pass': None,
'remote_user': hostvars.get('ansible_user'), # TODO
'password': (hostvars.get('ansible_ssh_pass') or
hostvars.get('ansible_password')),
'port': hostvars.get('ansible_port'),
'python_path': hostvars.get('ansible_python_interpreter'),
'private_key_file': (hostvars.get('ansible_ssh_private_key_file') or
hostvars.get('ansible_private_key_file')),
'mitogen_via': hostvars.get('mitogen_via'),
'mitogen_kind': hostvars.get('mitogen_kind'),
'mitogen_docker_path': hostvars.get('mitogen_docker_path'),
'mitogen_lxc_info_path': hostvars.get('mitogen_lxc_info_path'),
'mitogen_machinectl_path': hostvars.get('mitogen_machinctl_path'),
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
})
7 years ago
class Connection(ansible.plugins.connection.ConnectionBase):
#: mitogen.master.Broker for this worker.
broker = None
#: mitogen.master.Router for this worker.
router = None
#: mitogen.master.Context representing the parent Context, which is
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
#: presently always the connection multiplexer process.
parent = None
#: mitogen.master.Context connected to the target user account on the
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
#: target machine (i.e. via sudo).
7 years ago
context = 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
#: mitogen.master.Context connected to the fork parent process in the
#: target user account.
fork_context = None
#: Only sudo and su are supported for now.
become_methods = ['sudo', 'su']
#: Set to 'ansible_python_interpreter' by on_action_run().
python_path = None
#: Set to 'ansible_ssh_timeout' by on_action_run().
ansible_ssh_timeout = None
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
#: Set to 'mitogen_via' by on_action_run().
mitogen_via = None
#: Set to 'mitogen_kind' by on_action_run().
mitogen_kind = None
#: Set to 'mitogen_docker_path' by on_action_run().
mitogen_docker_path = None
#: Set to 'mitogen_lxc_info_path' by on_action_run().
mitogen_lxc_info_path = None
#: Set to 'mitogen_lxc_info_path' by on_action_run().
mitogen_machinectl_path = None
#: Set to 'mitogen_ssh_debug_level' by on_action_run().
mitogen_ssh_debug_level = None
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
#: Set to 'inventory_hostname' by on_action_run().
inventory_hostname = None
#: Set to 'hostvars' by on_action_run()
host_vars = None
#: Set after connection to the target context's home directory.
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
home_dir = None
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
def __init__(self, play_context, new_stdin, **kwargs):
assert ansible_mitogen.process.MuxProcess.unix_listener_path, (
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
'Mitogen connection types may only be instantiated '
'while the "mitogen" strategy is active.'
)
super(Connection, self).__init__(play_context, new_stdin)
def __del__(self):
"""
Ansible cannot be trusted to always call close() e.g. the synchronize
action constructs a local connection like this. So provide a destructor
in the hopes of catching these cases.
"""
# https://github.com/dw/mitogen/issues/140
self.close()
def on_action_run(self, task_vars):
"""
Invoked by ActionModuleMixin to indicate a new task is about to start
executing. We use the opportunity to grab relevant bits from the
task-specific data.
"""
self.ansible_ssh_timeout = task_vars.get('ansible_ssh_timeout',
C.DEFAULT_TIMEOUT)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
self.python_path = task_vars.get('ansible_python_interpreter',
'/usr/bin/python')
self.mitogen_via = task_vars.get('mitogen_via')
self.mitogen_kind = task_vars.get('mitogen_kind')
self.mitogen_docker_path = task_vars.get('mitogen_docker_path')
self.mitogen_lxc_info_path = task_vars.get('mitogen_lxc_info_path')
self.mitogen_machinectl_path = task_vars.get('mitogen_machinectl_path')
self.mitogen_ssh_debug_level = task_vars.get('mitogen_ssh_debug_level')
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
self.inventory_hostname = task_vars['inventory_hostname']
self.host_vars = task_vars['hostvars']
self.close(new_task=True)
@property
def homedir(self):
self._connect()
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.home_dir
7 years ago
@property
def connected(self):
return self.context is not None
7 years ago
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
def _config_from_via(self, via_spec):
become_user, _, inventory_name = via_spec.rpartition('@')
via_vars = self.host_vars[inventory_name]
if isinstance(via_vars, jinja2.runtime.Undefined):
raise ansible.errors.AnsibleConnectionFailure(
self.unknown_via_msg % (
self.mitogen_via,
inventory_name,
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
)
)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
return config_from_hostvars(
transport=via_vars.get('ansible_connection', 'ssh'),
inventory_name=inventory_name,
connection=self,
hostvars=via_vars,
become_user=become_user or None,
)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
unknown_via_msg = 'mitogen_via=%s of %s specifies an unknown hostname'
via_cycle_msg = 'mitogen_via=%s of %s creates a cycle (%s)'
def _stack_from_config(self, config, stack=(), seen_names=()):
if config['inventory_name'] in seen_names:
raise ansible.errors.AnsibleConnectionFailure(
self.via_cycle_msg % (
config['mitogen_via'],
config['inventory_name'],
' -> '.join(reversed(
seen_names + (config['inventory_name'],)
)),
)
)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
if config['mitogen_via']:
stack, seen_names = self._stack_from_config(
self._config_from_via(config['mitogen_via']),
stack=stack,
seen_names=seen_names + (config['inventory_name'],)
)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
stack += (CONNECTION_METHOD[config['transport']](config),)
if config['become']:
stack += (CONNECTION_METHOD[config['become_method']](config),)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
return stack, seen_names
def _connect(self):
"""
Establish a connection to the master process's UNIX listener socket,
constructing a mitogen.master.Router to communicate with the master,
and a mitogen.master.Context to represent it.
Depending on the original transport we should emulate, trigger one of
the _connect_*() service calls defined above to cause the master
process to establish the real connection on our behalf, or return a
reference to the existing one.
"""
if self.connected:
return
if not self.broker:
self.broker = mitogen.master.Broker()
self.router, self.parent = mitogen.unix.connect(
path=ansible_mitogen.process.MuxProcess.unix_listener_path,
broker=self.broker,
)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
stack, _ = self._stack_from_config(
config_from_play_context(
transport=self.transport,
inventory_name=self.inventory_hostname,
connection=self
)
)
dct = self.parent.call_service(
service_name='ansible_mitogen.services.ContextService',
method_name='get',
stack=mitogen.utils.cast(list(stack)),
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
)
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
if dct['msg']:
if dct['method_name'] in self.become_methods:
raise ansible.errors.AnsibleModuleError(dct['msg'])
raise ansible.errors.AnsibleConnectionFailure(dct['msg'])
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
self.context = dct['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
self.fork_context = dct['init_child_result']['fork_context']
self.home_dir = dct['init_child_result']['home_dir']
def close(self, new_task=False):
"""
Arrange for the mitogen.master.Router running in the worker to
gracefully shut down, and wait for shutdown to complete. Safe to call
multiple times.
"""
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
if self.context:
self.parent.call_service(
service_name='ansible_mitogen.services.ContextService',
method_name='put',
context=self.context
ansible: connection delegation v1 This implements the first edition of Connection Delegation, where delegating connection establishment is initially single-threaded. ansible_mitogen/strategy.py: ansible_mitogen/plugins/connection/*: Begin splitting connection.Connection into subclasses, exposing them directly as "mitogen_ssh", "mitogen_local", etc. connection types. This is far from removing strategy.py, but it's a tiny start. ansible_mitogen/connection.py: * config_from_play_context() and config_from_host_vars() build up a huge dictionary containing either more or less PlayContext contents, or our best attempt at reconstructing a host's connection config from its hostvars, where that config is not the current WorkerProcess target. They both produce the same format with the same keys, allowing remaining code to have a single input format. These dicts contain fields named after how Ansible refers to them, e.g. "sudo_exe". * _config_from_via() parses a basic connection specification like "username@inventory_name" into one of the aforementioned dicts. * _stack_from_config() produces a list of dicts describing the order in which (Mitogen) connections should be established, such that each element is proxied via= the previous element. The dicts produced by this function use Mitogen keyword arguments, the former di. These dicts contain fields named after how Mitogen refers to them, e.g. "sudo_path". * Pass the stack to ContextService, which is responsible for actual setup of the full chain. ansible_mitogen/services.py: Teach get() to walk the supplied stack, establishing each connection in turn, creating refounts for it before continuing. TODO: refcounting is broken in a variety of cases.
7 years ago
)
self.context = None
if self.broker and not new_task:
self.broker.shutdown()
self.broker.join()
self.broker = None
self.router = None
def call_async(self, func, *args, **kwargs):
"""
Start a function call to the target.
:returns:
mitogen.core.Receiver that receives the function call result.
"""
self._connect()
return self.context.call_async(func, *args, **kwargs)
def call(self, func, *args, **kwargs):
"""
Start and wait for completion of a function call in the target.
:raises mitogen.core.CallError:
The function call failed.
:returns:
Function return value.
"""
t0 = time.time()
try:
return self.call_async(func, *args, **kwargs).get().unpickle()
finally:
LOG.debug('Call took %d ms: %s%r', 1000 * (time.time() - t0),
func.__name__, args)
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 create_fork_child(self):
"""
Fork a new child off the target context. The actual fork occurs from
the 'virginal fork parent', which does not any Ansible modules prior to
fork, to avoid conflicts resulting from custom module_utils paths.
:returns:
mitogen.core.Context of the new child.
"""
return self.call(ansible_mitogen.target.create_fork_child)
def exec_command(self, cmd, in_data='', sudoable=True, mitogen_chdir=None):
"""
Implement exec_command() by calling the corresponding
ansible_mitogen.target function in the target.
:param str cmd:
Shell command to execute.
:param bytes in_data:
Data to supply on ``stdin`` of the process.
:returns:
(return code, stdout bytes, stderr bytes)
"""
emulate_tty = (not in_data and sudoable)
rc, stdout, stderr = self.call(
ansible_mitogen.target.exec_command,
cmd=mitogen.utils.cast(cmd),
in_data=mitogen.utils.cast(in_data),
chdir=mitogen_chdir,
emulate_tty=emulate_tty,
)
stderr += 'Shared connection to %s closed.%s' % (
self._play_context.remote_addr,
('\r\n' if emulate_tty else '\n'),
)
return rc, stdout, stderr
7 years ago
def fetch_file(self, in_path, out_path):
"""
Implement fetch_file() by calling the corresponding
ansible_mitogen.target function in the target.
:param str in_path:
Remote filesystem path to read.
:param str out_path:
Local filesystem path to write.
"""
output = self.call(ansible_mitogen.target.read_path,
mitogen.utils.cast(in_path))
ansible_mitogen.target.write_path(out_path, output)
7 years ago
def put_data(self, out_path, data, mode=None, utimes=None):
"""
Implement put_file() by caling the corresponding
ansible_mitogen.target function in the target.
:param str out_path:
Remote filesystem path to write.
:param byte data:
File contents to put.
"""
self.call(ansible_mitogen.target.write_path,
mitogen.utils.cast(out_path),
mitogen.core.Blob(data),
mode=mode,
utimes=utimes)
7 years ago
def put_file(self, in_path, out_path):
"""
Implement put_file() by streamily transferring the file via
FileService.
:param str in_path:
Local filesystem path to read.
:param str out_path:
Remote filesystem path to write.
"""
st = os.stat(in_path)
if not stat.S_ISREG(st.st_mode):
raise IOError('%r is not a regular file.' % (in_path,))
# If the file is sufficiently small, just ship it in the argument list
# rather than introducing an extra RTT for the child to request it from
# FileService.
if st.st_size <= 32768:
fp = open(in_path, 'rb')
try:
s = fp.read(32769)
finally:
fp.close()
# Ensure file was not growing during call.
if len(s) == st.st_size:
return self.put_data(out_path, s, mode=st.st_mode,
utimes=(st.st_atime, st.st_mtime))
self.parent.call_service(
service_name='mitogen.service.FileService',
method_name='register',
path=mitogen.utils.cast(in_path)
)
self.call(
ansible_mitogen.target.transfer_file,
context=self.parent,
in_path=in_path,
out_path=out_path
)