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/process.py

284 lines
10 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
import atexit
import errno
import logging
import os
import signal
import socket
import sys
import time
try:
import faulthandler
except ImportError:
faulthandler = None
import mitogen
import mitogen.core
import mitogen.debug
import mitogen.master
import mitogen.parent
import mitogen.service
import mitogen.unix
import mitogen.utils
import ansible.constants as C
import ansible_mitogen.logging
import ansible_mitogen.services
from mitogen.core import b
LOG = logging.getLogger(__name__)
def clean_shutdown(sock):
"""
Shut the write end of `sock`, causing `recv` in the worker process to wake
up with a 0-byte read and initiate mux process exit, then wait for a 0-byte
read from the read end, which will occur after the the child closes the
descriptor on exit.
This is done using :mod:`atexit` since Ansible lacks any more sensible hook
to run code during exit, and unless some synchronization exists with
MuxProcess, debug logs may appear on the user's terminal *after* the prompt
has been printed.
"""
sock.shutdown(socket.SHUT_WR)
sock.recv(1)
def getenv_int(key, default=0):
"""
Get an integer-valued environment variable `key`, if it exists and parses
as an integer, otherwise return `default`.
"""
try:
return int(os.environ.get(key, str(default)))
except ValueError:
return default
def setup_gil():
"""
Set extremely long GIL release interval to let threads naturally progress
through CPU-heavy sequences without forcing the wake of another thread that
may contend trying to run the same CPU-heavy code. For the new-style work,
this drops runtime ~33% and involuntary context switches by >80%,
essentially making threads cooperatively scheduled.
"""
try:
# Python 2.
sys.setcheckinterval(100000)
except AttributeError:
pass
try:
# Python 3.
sys.setswitchinterval(10)
except AttributeError:
pass
class MuxProcess(object):
"""
Implement a subprocess forked from the Ansible top-level, as a safe place
to contain the Mitogen IO multiplexer thread, keeping its use of the
logging package (and the logging package's heavy use of locks) far away
from the clutches of os.fork(), which is used continuously by the
multiprocessing package in the top-level process.
The problem with running the multiplexer in that process is that should the
multiplexer thread be in the process of emitting a log entry (and holding
its lock) at the point of fork, in the child, the first attempt to log any
log entry using the same handler will deadlock the child, as in the memory
image the child received, the lock will always be marked held.
See https://bugs.python.org/issue6721 for a thorough description of the
class of problems this worker is intended to avoid.
"""
#: In the top-level process, this references one end of a socketpair(),
#: which the MuxProcess blocks reading from in order to determine when
#: the master process dies. Once the read returns, the MuxProcess will
#: begin shutting itself down.
worker_sock = None
#: In the worker process, this references the other end of
#: :py:attr:`worker_sock`.
child_sock = None
#: In the top-level process, this is the PID of the single MuxProcess
#: that was spawned.
worker_pid = None
#: A copy of :data:`os.environ` at the time the multiplexer process was
#: started. It's used by mitogen_local.py to find changes made to the
#: top-level environment (e.g. vars plugins -- issue #297) that must be
#: applied to locally executed commands and modules.
original_env = None
#: In both processes, this is the temporary UNIX socket used for
#: forked WorkerProcesses to contact the MuxProcess
unix_listener_path = None
#: Singleton.
_instance = None
@classmethod
def start(cls, _init_logging=True):
"""
Arrange for the subprocess to be started, if it is not already running.
The parent process picks a UNIX socket path the child will use prior to
fork, creates a socketpair used essentially as a semaphore, then blocks
waiting for the child to indicate the UNIX socket is ready for use.
:param bool _init_logging:
For testing, if :data:`False`, don't initialize logging.
"""
if cls.worker_sock is not None:
return
if faulthandler is not None:
faulthandler.enable()
setup_gil()
cls.unix_listener_path = mitogen.unix.make_socket_path()
cls.worker_sock, cls.child_sock = socket.socketpair()
atexit.register(lambda: clean_shutdown(cls.worker_sock))
mitogen.core.set_cloexec(cls.worker_sock.fileno())
mitogen.core.set_cloexec(cls.child_sock.fileno())
if os.environ.get('MITOGEN_PROFILING'):
mitogen.core.enable_profiling()
cls.original_env = dict(os.environ)
cls.child_pid = os.fork()
if _init_logging:
ansible_mitogen.logging.setup()
if cls.child_pid:
cls.child_sock.close()
cls.child_sock = None
mitogen.core.io_op(cls.worker_sock.recv, 1)
else:
cls.worker_sock.close()
cls.worker_sock = None
self = cls()
self.worker_main()
sys.exit()
def worker_main(self):
"""
The main function of for the mux process: setup the Mitogen broker
thread and ansible_mitogen services, then sleep waiting for the socket
connected to the parent to be closed (indicating the parent has died).
"""
self._setup_master()
self._setup_services()
# Let the parent know our listening socket is ready.
mitogen.core.io_op(self.child_sock.send, b('1'))
# Block until the socket is closed, which happens on parent exit.
mitogen.core.io_op(self.child_sock.recv, 1)
def _enable_router_debug(self):
if 'MITOGEN_ROUTER_DEBUG' in os.environ:
self.router.enable_debug()
def _enable_stack_dumps(self):
secs = getenv_int('MITOGEN_DUMP_THREAD_STACKS', default=0)
if secs:
mitogen.debug.dump_to_logger(secs=secs)
def _setup_master(self):
"""
Construct a Router, Broker, and mitogen.unix listener
"""
self.router = mitogen.master.Router(max_message_size=4096 * 1048576)
self.router.responder.whitelist_prefix('ansible')
self.router.responder.whitelist_prefix('ansible_mitogen')
mitogen.core.listen(self.router.broker, 'shutdown', self.on_broker_shutdown)
mitogen.core.listen(self.router.broker, 'exit', self.on_broker_exit)
self.listener = mitogen.unix.Listener(
router=self.router,
path=self.unix_listener_path,
backlog=C.DEFAULT_FORKS,
)
self._enable_router_debug()
self._enable_stack_dumps()
def _setup_services(self):
"""
Construct a ContextService and a thread to service requests for it
arriving from worker processes.
"""
self.pool = mitogen.service.Pool(
router=self.router,
services=[
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.service.FileService(router=self.router),
mitogen.service.PushFileService(router=self.router),
ansible_mitogen.services.ContextService(self.router),
issue #186: rework async/forked tasks again. The controller must know the ID of the forked child in order to propagate dependencies to it, so forking+starting the module run cannot happen entirely on the target, without some additional mechanism to wait-and-repropagate the deps as they arrive on the target. Rework things so that init_child() also handles starting the fork parent, and returns it along with the context's home directory in a single round trip. Now master knows the identity of the fork parent, it can directly create fork children and call run_module_async() in them. This necessitates 2 roundtrips to start an asynchronous task. This whole thing sucks and entirely needs simplified, but for now things almost work, so keeping it. connection.py: * Expect ContextService to return the entire dict return value of init_child(). Store the fork_contxt from the return value. planner.py: * Rework Planner to store the invocation as an instance attribute, to simplify method calls. * Add Planner.get_push_files() and Planner.get_module_deps(). * Add _propagate_deps() which takes a Planner and ensures the deps it describes are sent to a (non forked or forked) context. * Move async task logic out of target.py and into invoke() / _invoke_*(). process.py: * Services no longer need references to each other. planner.py handles sending module deps with one extra RPC. services.py: * Return "init_child_result" key instead of simple "home_dir" key. * Get rid of dep propagation from ModuleDepService, it lives in planner.py now. target.py: * Get rid of async task start logic, lives in planner.py now.
7 years ago
ansible_mitogen.services.ModuleDepService(self.router),
],
size=getenv_int('MITOGEN_POOL_SIZE', default=16),
)
LOG.debug('Service pool configured: size=%d', self.pool.size)
def on_broker_shutdown(self):
"""
Respond to broker shutdown by beginning service pool shutdown. Do not
join on the pool yet, since that would block the broker thread which
then cannot clean up pending handlers, which is required for the
threads to exit gracefully.
"""
self.pool.stop(join=False)
try:
os.unlink(self.listener.path)
except OSError as e:
# Prevent a shutdown race with the parent process.
if e.args[0] != errno.ENOENT:
raise
def on_broker_exit(self):
"""
Respond to the broker thread about to exit by sending SIGTERM to
ourself. In future this should gracefully join the pool, but TERM is
fine for now.
"""
if os.environ.get('MITOGEN_PROFILING'):
# TODO: avoid killing pool threads before they have written their
# .pstats. Really shouldn't be using kill() here at all, but hard
# to guarantee services can always be unblocked during shutdown.
time.sleep(1)
os.kill(os.getpid(), signal.SIGTERM)