def _inner():
if initial_delay:
greenthread.sleep(initial_delay)
try:
while self._running:
start = timeutils.utcnow()
self.f(*self.args, **self.kw)
end = timeutils.utcnow()
if not self._running:
break
delay = interval - timeutils.delta_seconds(start, end)
if delay <= 0:
LOG.warn(
_('task run outlasted interval by %s sec') %
-delay)
greenthread.sleep(delay if delay > 0 else 0)
except LoopingCallDone as e:
self.stop()
done.send(e.retvalue)
except Exception:
LOG.exception(_('in fixed duration looping call'))
done.send_exception(*sys.exc_info())
return
else:
done.send(True)
def reconnect(self):
"""Handles reconnecting and re-establishing sessions and queues"""
attempt = 0
delay = 1
while True:
# Close the session if necessary
if self.connection.opened():
try:
self.connection.close()
except qpid_exceptions.ConnectionError:
pass
broker = self.brokers[attempt % len(self.brokers)]
attempt += 1
try:
self.connection_create(broker)
self.connection.open()
except qpid_exceptions.ConnectionError, e:
msg_dict = dict(e=e, delay=delay)
msg = _("Unable to connect to AMQP server: %(e)s. "
"Sleeping %(delay)s seconds") % msg_dict
LOG.error(msg)
time.sleep(delay)
delay = min(2 * delay, 60)
else:
LOG.info(_('Connected to AMQP server on %s'), broker)
break
def _multi_send(method, context, topic, msg, timeout=None,
envelope=False, _msg_id=None):
"""Wraps the sending of messages.
Dispatches to the matchmaker and sends message to all relevant hosts.
"""
conf = CONF
LOG.debug(_("%(msg)s") % {'msg': ' '.join(map(pformat, (topic, msg)))})
queues = _get_matchmaker().queues(topic)
LOG.debug(_("Sending message(s) to: %s"), queues)
# Don't stack if we have no matchmaker results
if not queues:
LOG.warn(_("No matchmaker results. Not casting."))
# While not strictly a timeout, callers know how to handle
# this exception and a timeout isn't too big a lie.
raise rpc_common.Timeout(_("No match from matchmaker."))
# This supports brokerless fanout (addresses > 1)
for queue in queues:
(_topic, ip_addr) = queue
_addr = "tcp://%s:%s" % (ip_addr, conf.rpc_zmq_port)
if method.__name__ == '_cast':
eventlet.spawn_n(method, _addr, context,
_topic, msg, timeout, envelope,
_msg_id)
return
return method(_addr, context, _topic, msg, timeout,
envelope)
def consume_in_thread(self):
"""Runs the ZmqProxy service."""
ipc_dir = CONF.rpc_zmq_ipc_dir
consume_in = "tcp://%s:%s" % \
(CONF.rpc_zmq_bind_address,
CONF.rpc_zmq_port)
consumption_proxy = InternalContext(None)
try:
os.makedirs(ipc_dir)
except os.error:
if not os.path.isdir(ipc_dir):
with excutils.save_and_reraise_exception():
LOG.error(
_("Required IPC directory does not exist at"
" %s") % (ipc_dir, ))
try:
self.register(consumption_proxy, consume_in, zmq.PULL)
except zmq.ZMQError:
if os.access(ipc_dir, os.X_OK):
with excutils.save_and_reraise_exception():
LOG.error(
_("Permission denied to IPC directory at"
" %s") % (ipc_dir, ))
with excutils.save_and_reraise_exception():
LOG.error(
_("Could not create ZeroMQ receiver daemon. "
"Socket may already be in use."))
super(ZmqProxy, self).consume_in_thread()
def create_consumer(self, topic, proxy, fanout=False):
# Register with matchmaker.
_get_matchmaker().register(topic, CONF.rpc_zmq_host)
# Subscription scenarios
if fanout:
sock_type = zmq.SUB
subscribe = ('', fanout)[type(fanout) == str]
topic = 'fanout~' + topic.split('.', 1)[0]
else:
sock_type = zmq.PULL
subscribe = None
topic = '.'.join((topic.split('.', 1)[0], CONF.rpc_zmq_host))
if topic in self.topics:
LOG.info(_("Skipping topic registration. Already registered."))
return
# Receive messages from (local) proxy
inaddr = "ipc://%s/zmq_topic_%s" % \
(CONF.rpc_zmq_ipc_dir, topic)
LOG.debug(_("Consumer is a zmq.%s"),
['PULL', 'SUB'][sock_type == zmq.SUB])
self.reactor.register(proxy, inaddr, sock_type,
subscribe=subscribe, in_bind=False)
self.topics.append(topic)
def publisher(waiter):
LOG.info(_("Creating proxy for topic: %s"), topic)
try:
# The topic is received over the network,
# don't trust this input.
if self.badchars.search(topic) is not None:
emsg = _("Topic contained dangerous characters.")
LOG.warn(emsg)
raise RPCException(emsg)
out_sock = ZmqSocket("ipc://%s/zmq_topic_%s" %
(ipc_dir, topic),
sock_type,
bind=True)
except RPCException:
waiter.send_exception(*sys.exc_info())
return
self.topic_proxy[topic] = eventlet.queue.LightQueue(
CONF.rpc_zmq_topic_backlog)
self.sockets.append(out_sock)
# It takes some time for a pub socket to open,
# before we can have any faith in doing a send() to it.
if sock_type == zmq.PUB:
eventlet.sleep(.5)
waiter.send(True)
while (True):
data = self.topic_proxy[topic].get()
out_sock.send(data, copy=False)
def _get_response(self, ctx, proxy, topic, data):
"""Process a curried message and cast the result to topic."""
LOG.debug(_("Running func with context: %s"), ctx.to_dict())
data.setdefault('version', None)
data.setdefault('args', {})
try:
result = proxy.dispatch(ctx, data['version'], data['method'],
data.get('namespace'), **data['args'])
return ConsumerBase.normalize_reply(result, ctx.replies)
except greenlet.GreenletExit:
# ignore these since they are just from shutdowns
pass
except rpc_common.ClientException as e:
LOG.debug(
_("Expected exception during message handling (%s)") %
e._exc_info[1])
return {
'exc':
rpc_common.serialize_remote_exception(e._exc_info,
log_failure=False)
}
except Exception:
LOG.error(_("Exception during message handling"))
return {
'exc': rpc_common.serialize_remote_exception(sys.exc_info())
}
def _parse_check(rule):
"""
Parse a single base check rule into an appropriate Check object.
"""
# Handle the special checks
if rule == '!':
return FalseCheck()
elif rule == '@':
return TrueCheck()
try:
kind, match = rule.split(':', 1)
except Exception:
LOG.exception(_("Failed to understand rule %(rule)s") % locals())
# If the rule is invalid, we'll fail closed
return FalseCheck()
# Find what implements the check
if kind in _checks:
return _checks[kind](kind, match)
elif None in _checks:
return _checks[None](kind, match)
else:
LOG.error(_("No handler for matches of kind %s") % kind)
return FalseCheck()
def reconnect(self):
"""Handles reconnecting and re-establishing sessions and queues"""
attempt = 0
delay = 1
while True:
# Close the session if necessary
if self.connection.opened():
try:
self.connection.close()
except qpid_exceptions.ConnectionError:
pass
broker = self.brokers[attempt % len(self.brokers)]
attempt += 1
try:
self.connection_create(broker)
self.connection.open()
except qpid_exceptions.ConnectionError, e:
msg_dict = dict(e=e, delay=delay)
msg = _("Unable to connect to AMQP server: %(e)s. "
"Sleeping %(delay)s seconds") % msg_dict
LOG.error(msg)
time.sleep(delay)
delay = min(2 * delay, 60)
else:
LOG.info(_('Connected to AMQP server on %s'), broker)
break
def wait(self):
"""Loop waiting on children to die and respawning as necessary"""
LOG.debug(_('Full set of CONF:'))
CONF.log_opt_values(LOG, std_logging.DEBUG)
while self.running:
wrap = self._wait_child()
if not wrap:
# Yield to other threads if no children have exited
# Sleep for a short time to avoid excessive CPU usage
# (see bug #1095346)
eventlet.greenthread.sleep(.01)
continue
while self.running and len(wrap.children) < wrap.workers:
self._start_child(wrap)
if self.sigcaught:
signame = {signal.SIGTERM: 'SIGTERM',
signal.SIGINT: 'SIGINT'}[self.sigcaught]
LOG.info(_('Caught %s, stopping children'), signame)
for pid in self.children:
try:
os.kill(pid, signal.SIGTERM)
except OSError as exc:
if exc.errno != errno.ESRCH:
raise
# Wait for children to die
if self.children:
LOG.info(_('Waiting on %d children to exit'), len(self.children))
while self.children:
self._wait_child()
def _process_data(self, ctxt, version, method, namespace, args):
"""Process a message in a new thread.
If the proxy object we have has a dispatch method
(see rpc.dispatcher.RpcDispatcher), pass it the version,
method, and args and let it dispatch as appropriate. If not, use
the old behavior of magically calling the specified method on the
proxy we have here.
"""
ctxt.update_store()
try:
rval = self.proxy.dispatch(ctxt, version, method, namespace,
**args)
# Check if the result was a generator
if inspect.isgenerator(rval):
for x in rval:
ctxt.reply(x, None, connection_pool=self.connection_pool)
else:
ctxt.reply(rval, None, connection_pool=self.connection_pool)
# This final None tells multicall that it is done.
ctxt.reply(ending=True, connection_pool=self.connection_pool)
except rpc_common.ClientException as e:
LOG.debug(_('Expected exception during message handling (%s)') %
e._exc_info[1])
ctxt.reply(None, e._exc_info,
connection_pool=self.connection_pool,
log_failure=False)
except Exception:
# sys.exc_info() is deleted by LOG.exception().
exc_info = sys.exc_info()
LOG.error(_('Exception during message handling'),
exc_info=exc_info)
ctxt.reply(None, exc_info, connection_pool=self.connection_pool)
def _wait_child(self):
try:
# Don't block if no child processes have exited
pid, status = os.waitpid(0, os.WNOHANG)
if not pid:
return None
except OSError as exc:
if exc.errno not in (errno.EINTR, errno.ECHILD):
raise
return None
if os.WIFSIGNALED(status):
sig = os.WTERMSIG(status)
LOG.info(_('Child %(pid)d killed by signal %(sig)d'),
dict(pid=pid, sig=sig))
else:
code = os.WEXITSTATUS(status)
LOG.info(_('Child %(pid)s exited with status %(code)d'),
dict(pid=pid, code=code))
if pid not in self.children:
LOG.warning(_('pid %d not in child list'), pid)
return None
wrap = self.children.pop(pid)
wrap.children.remove(pid)
return wrap
def consume(self, sock):
#TODO(ewindisch): use zero-copy (i.e. references, not copying)
data = sock.recv()
LOG.debug(_("CONSUMER RECEIVED DATA: %s"), data)
proxy = self.proxies[sock]
if data[2] == 'cast': # Legacy protocol
packenv = data[3]
ctx, msg = _deserialize(packenv)
request = rpc_common.deserialize_msg(msg)
ctx = RpcContext.unmarshal(ctx)
elif data[2] == 'impl_zmq_v2':
packenv = data[4:]
msg = unflatten_envelope(packenv)
request = rpc_common.deserialize_msg(msg)
# Unmarshal only after verifying the message.
ctx = RpcContext.unmarshal(data[3])
else:
LOG.error(_("ZMQ Envelope version unsupported or unknown."))
return
self.pool.spawn_n(self.process, proxy, ctx, request)
def consume_in_thread(self):
"""Runs the ZmqProxy service."""
ipc_dir = CONF.rpc_zmq_ipc_dir
consume_in = "tcp://%s:%s" % \
(CONF.rpc_zmq_bind_address,
CONF.rpc_zmq_port)
consumption_proxy = InternalContext(None)
try:
os.makedirs(ipc_dir)
except os.error:
if not os.path.isdir(ipc_dir):
with excutils.save_and_reraise_exception():
LOG.error(_("Required IPC directory does not exist at"
" %s") % (ipc_dir, ))
try:
self.register(consumption_proxy,
consume_in,
zmq.PULL)
except zmq.ZMQError:
if os.access(ipc_dir, os.X_OK):
with excutils.save_and_reraise_exception():
LOG.error(_("Permission denied to IPC directory at"
" %s") % (ipc_dir, ))
with excutils.save_and_reraise_exception():
LOG.error(_("Could not create ZeroMQ receiver daemon. "
"Socket may already be in use."))
super(ZmqProxy, self).consume_in_thread()
def run_periodic_tasks(self, context, raise_on_error=False):
"""Tasks to be run at a periodic interval."""
for task_name, task in self._periodic_tasks:
full_task_name = '.'.join([self.__class__.__name__, task_name])
ticks_to_skip = self._ticks_to_skip[task_name]
if ticks_to_skip > 0:
LOG.debug(
_("Skipping %(full_task_name)s, %(ticks_to_skip)s"
" ticks left until next run"),
dict(full_task_name=full_task_name,
ticks_to_skip=ticks_to_skip))
self._ticks_to_skip[task_name] -= 1
continue
self._ticks_to_skip[task_name] = task._ticks_between_runs
LOG.debug(_("Running periodic task %(full_task_name)s"),
dict(full_task_name=full_task_name))
try:
task(self, context)
except Exception as e:
if raise_on_error:
raise
LOG.exception(_("Error during %(full_task_name)s:"
" %(e)s"),
dict(e=e, full_task_name=full_task_name))
def ssh_execute(ssh, cmd, process_input=None, addl_env=None, check_exit_code=True):
LOG.debug("Running cmd (SSH): %s", cmd)
if addl_env:
raise InvalidArgumentError(_("Environment not supported over SSH"))
if process_input:
# This is (probably) fixable if we need it...
raise InvalidArgumentError(_("process_input not supported over SSH"))
stdin_stream, stdout_stream, stderr_stream = ssh.exec_command(cmd)
channel = stdout_stream.channel
# NOTE(justinsb): This seems suspicious...
# ...other SSH clients have buffering issues with this approach
stdout = stdout_stream.read()
stderr = stderr_stream.read()
stdin_stream.close()
exit_status = channel.recv_exit_status()
# exit_status == -1 if no exit code was returned
if exit_status != -1:
LOG.debug("Result was %s" % exit_status)
if check_exit_code and exit_status != 0:
raise ProcessExecutionError(exit_code=exit_status, stdout=stdout, stderr=stderr, cmd=cmd)
return (stdout, stderr)
def wait(self):
"""Loop waiting on children to die and respawning as necessary."""
LOG.debug(_('Full set of CONF:'))
CONF.log_opt_values(LOG, std_logging.DEBUG)
while True:
self.handle_signal()
self._respawn_children()
if self.sigcaught:
signame = _signo_to_signame(self.sigcaught)
LOG.info(_('Caught %s, stopping children'), signame)
if not _is_sighup_and_daemon(self.sigcaught):
break
for pid in self.children:
os.kill(pid, signal.SIGHUP)
self.running = True
self.sigcaught = None
for pid in self.children:
try:
os.kill(pid, signal.SIGTERM)
except OSError as exc:
if exc.errno != errno.ESRCH:
raise
# Wait for children to die
if self.children:
LOG.info(_('Waiting on %d children to exit'), len(self.children))
while self.children:
self._wait_child()
def to_bytes(text, default=0):
"""Converts a string into an integer of bytes.
Looks at the last characters of the text to determine
what conversion is needed to turn the input text into a byte number.
Supports "B, K(B), M(B), G(B), and T(B)". (case insensitive)
:param text: String input for bytes size conversion.
:param default: Default return value when text is blank.
"""
match = BYTE_REGEX.search(text)
if match:
magnitude = int(match.group(1))
mult_key_org = match.group(2)
if not mult_key_org:
return magnitude
elif text:
msg = _('Invalid string format: %s') % text
raise TypeError(msg)
else:
return default
mult_key = mult_key_org.lower().replace('b', '', 1)
multiplier = BYTE_MULTIPLIERS.get(mult_key)
if multiplier is None:
msg = _('Unknown byte multiplier: %s') % mult_key_org
raise TypeError(msg)
return magnitude * multiplier
def _wait_for_exit_or_signal(self, ready_callback=None):
status = None
signo = 0
LOG.debug(_('Full set of CONF:'))
CONF.log_opt_values(LOG, std_logging.DEBUG)
try:
if ready_callback:
ready_callback()
super(ServiceLauncher, self).wait()
except SignalExit as exc:
signame = _signo_to_signame(exc.signo)
LOG.info(_('Caught %s, exiting'), signame)
status = exc.code
signo = exc.signo
except SystemExit as exc:
status = exc.code
finally:
self.stop()
if rpc:
try:
rpc.cleanup()
except Exception:
# We're shutting down, so it doesn't matter at this point.
LOG.exception(_('Exception during rpc cleanup.'))
return status, signo
def _start_child(self, wrap):
if len(wrap.forktimes) > wrap.workers:
# Limit ourselves to one process a second (over the period of
# number of workers * 1 second). This will allow workers to
# start up quickly but ensure we don't fork off children that
# die instantly too quickly.
if time.time() - wrap.forktimes[0] < wrap.workers:
LOG.info(_('Forking too fast, sleeping'))
time.sleep(1)
wrap.forktimes.pop(0)
wrap.forktimes.append(time.time())
pid = os.fork()
if pid == 0:
launcher = self._child_process(wrap.service)
while True:
self._child_process_handle_signal()
status, signo = self._child_wait_for_exit_or_signal(launcher)
if not _is_sighup_and_daemon(signo):
break
launcher.restart()
os._exit(status)
LOG.info(_('Started child %d'), pid)
wrap.children.add(pid)
self.children[pid] = wrap
return pid
def _multi_send(method,
context,
topic,
msg,
timeout=None,
envelope=False,
_msg_id=None):
"""
Wraps the sending of messages,
dispatches to the matchmaker and sends
message to all relevant hosts.
"""
conf = CONF
LOG.debug(_("%(msg)s") % {'msg': ' '.join(map(pformat, (topic, msg)))})
queues = _get_matchmaker().queues(topic)
LOG.debug(_("Sending message(s) to: %s"), queues)
# Don't stack if we have no matchmaker results
if len(queues) == 0:
LOG.warn(_("No matchmaker results. Not casting."))
# While not strictly a timeout, callers know how to handle
# this exception and a timeout isn't too big a lie.
raise rpc_common.Timeout(_("No match from matchmaker."))
# This supports brokerless fanout (addresses > 1)
for queue in queues:
(_topic, ip_addr) = queue
_addr = "tcp://%s:%s" % (ip_addr, conf.rpc_zmq_port)
if method.__name__ == '_cast':
eventlet.spawn_n(method, _addr, context, _topic, msg, timeout,
envelope, _msg_id)
return
return method(_addr, context, _topic, msg, timeout, envelope)
def create_consumer(self, topic, proxy, fanout=False):
# Register with matchmaker.
_get_matchmaker().register(topic, CONF.rpc_zmq_host)
# Subscription scenarios
if fanout:
sock_type = zmq.SUB
subscribe = ('', fanout)[type(fanout) == str]
topic = 'fanout~' + topic.split('.', 1)[0]
else:
sock_type = zmq.PULL
subscribe = None
topic = '.'.join((topic.split('.', 1)[0], CONF.rpc_zmq_host))
if topic in self.topics:
LOG.info(_("Skipping topic registration. Already registered."))
return
# Receive messages from (local) proxy
inaddr = "ipc://%s/zmq_topic_%s" % \
(CONF.rpc_zmq_ipc_dir, topic)
LOG.debug(_("Consumer is a zmq.%s"), ['PULL',
'SUB'][sock_type == zmq.SUB])
self.reactor.register(proxy,
inaddr,
sock_type,
subscribe=subscribe,
in_bind=False)
self.topics.append(topic)
def consume(self, sock):
#TODO(ewindisch): use zero-copy (i.e. references, not copying)
data = sock.recv()
LOG.debug(_("CONSUMER RECEIVED DATA: %s"), data)
if sock in self.mapping:
LOG.debug(_("ROUTER RELAY-OUT %(data)s") % {'data': data})
self.mapping[sock].send(data)
return
proxy = self.proxies[sock]
if data[2] == 'cast': # Legacy protocol
packenv = data[3]
ctx, msg = _deserialize(packenv)
request = rpc_common.deserialize_msg(msg)
ctx = RpcContext.unmarshal(ctx)
elif data[2] == 'impl_zmq_v2':
packenv = data[4:]
msg = unflatten_envelope(packenv)
request = rpc_common.deserialize_msg(msg)
# Unmarshal only after verifying the message.
ctx = RpcContext.unmarshal(data[3])
else:
LOG.error(_("ZMQ Envelope version unsupported or unknown."))
return
self.pool.spawn_n(self.process, proxy, ctx, request)
def __call__(self, message_data):
"""Consumer callback to call a method on a proxy object.
Parses the message for validity and fires off a thread to call the
proxy object method.
Message data should be a dictionary with two keys:
method: string representing the method to call
args: dictionary of arg: value
Example: {'method': 'echo', 'args': {'value': 42}}
"""
# It is important to clear the context here, because at this point
# the previous context is stored in local.store.context
if hasattr(local.store, 'context'):
del local.store.context
rpc_common._safe_log(LOG.debug, _('received %s'), message_data)
self.msg_id_cache.check_duplicate_message(message_data)
ctxt = unpack_context(self.conf, message_data)
method = message_data.get('method')
args = message_data.get('args', {})
version = message_data.get('version')
namespace = message_data.get('namespace')
if not method:
LOG.warn(_('no method for message: %s') % message_data)
ctxt.reply(_('No method for message: %s') % message_data,
connection_pool=self.connection_pool)
return
self.pool.spawn_n(self._process_data, ctxt, version, method,
namespace, args)
def _inner():
if initial_delay:
greenthread.sleep(initial_delay)
try:
while self._running:
idle = self.f(*self.args, **self.kw)
if not self._running:
break
if periodic_interval_max is not None:
idle = min(idle, periodic_interval_max)
LOG.debug(
_('Dynamic looping call sleeping for %.02f '
'seconds'), idle)
greenthread.sleep(idle)
except LoopingCallDone as e:
self.stop()
done.send(e.retvalue)
except Exception:
LOG.exception(_('in dynamic looping call'))
done.send_exception(*sys.exc_info())
return
else:
done.send(True)
def run_periodic_tasks(self, context, raise_on_error=False):
"""Tasks to be run at a periodic interval."""
for task_name, task in self._periodic_tasks:
full_task_name = '.'.join([self.__class__.__name__, task_name])
ticks_to_skip = self._ticks_to_skip[task_name]
if ticks_to_skip > 0:
LOG.debug(_("Skipping %(full_task_name)s, %(ticks_to_skip)s"
" ticks left until next run"),
dict(full_task_name=full_task_name,
ticks_to_skip=ticks_to_skip))
self._ticks_to_skip[task_name] -= 1
continue
self._ticks_to_skip[task_name] = task._ticks_between_runs
LOG.debug(_("Running periodic task %(full_task_name)s"),
dict(full_task_name=full_task_name))
try:
task(self, context)
except Exception as e:
if raise_on_error:
raise
LOG.exception(_("Error during %(full_task_name)s:"
" %(e)s"),
dict(e=e, full_task_name=full_task_name))
def _connect(self, params):
"""Connect to rabbit. Re-establish any queues that may have
been declared before if we are reconnecting. Exceptions should
be handled by the caller.
"""
if self.connection:
LOG.info(_("Reconnecting to AMQP server on "
"%(hostname)s:%(port)d") % params)
try:
self.connection.release()
except self.connection_errors:
pass
# Setting this in case the next statement fails, though
# it shouldn't be doing any network operations, yet.
self.connection = None
self.connection = kombu.connection.BrokerConnection(**params)
self.connection_errors = self.connection.connection_errors
if self.memory_transport:
# Kludge to speed up tests.
self.connection.transport.polling_interval = 0.0
self.consumer_num = itertools.count(1)
self.connection.connect()
self.channel = self.connection.channel()
# work around 'memory' transport bug in 1.1.3
if self.memory_transport:
self.channel._new_queue('ae.undeliver')
for consumer in self.consumers:
consumer.reconnect(self.channel)
LOG.info(_('Connected to AMQP server on %(hostname)s:%(port)d') %
params)
def wait(self):
"""Loop waiting on children to die and respawning as necessary."""
LOG.debug(_('Full set of CONF:'))
CONF.log_opt_values(LOG, std_logging.DEBUG)
while True:
self.handle_signal()
self._respawn_children()
if self.sigcaught:
signame = _signo_to_signame(self.sigcaught)
LOG.info(_('Caught %s, stopping children'), signame)
if not _is_sighup_and_daemon(self.sigcaught):
break
for pid in self.children:
os.kill(pid, signal.SIGHUP)
self.running = True
self.sigcaught = None
for pid in self.children:
try:
os.kill(pid, signal.SIGTERM)
except OSError as exc:
if exc.errno != errno.ESRCH:
raise
# Wait for children to die
if self.children:
LOG.info(_('Waiting on %d children to exit'), len(self.children))
while self.children:
self._wait_child()
def __init__(self, addr, zmq_type, bind=True, subscribe=None):
self.sock = _get_ctxt().socket(zmq_type)
self.addr = addr
self.type = zmq_type
self.subscriptions = []
# Support failures on sending/receiving on wrong socket type.
self.can_recv = zmq_type in (zmq.PULL, zmq.SUB)
self.can_send = zmq_type in (zmq.PUSH, zmq.PUB)
self.can_sub = zmq_type in (zmq.SUB, )
# Support list, str, & None for subscribe arg (cast to list)
do_sub = {
list: subscribe,
str: [subscribe],
type(None): []
}[type(subscribe)]
for f in do_sub:
self.subscribe(f)
str_data = {'addr': addr, 'type': self.socket_s(),
'subscribe': subscribe, 'bind': bind}
LOG.debug(_("Connecting to %(addr)s with %(type)s"), str_data)
LOG.debug(_("-> Subscribed to %(subscribe)s"), str_data)
LOG.debug(_("-> bind: %(bind)s"), str_data)
try:
if bind:
self.sock.bind(addr)
else:
self.sock.connect(addr)
except Exception:
raise RPCException(_("Could not open socket."))
def publisher(waiter):
LOG.info(_("Creating proxy for topic: %s"), topic)
try:
# The topic is received over the network,
# don't trust this input.
if self.badchars.search(topic) is not None:
emsg = _("Topic contained dangerous characters.")
LOG.warn(emsg)
raise RPCException(emsg)
out_sock = ZmqSocket("ipc://%s/zmq_topic_%s" %
(ipc_dir, topic),
sock_type, bind=True)
except RPCException:
waiter.send_exception(*sys.exc_info())
return
self.topic_proxy[topic] = eventlet.queue.LightQueue(
CONF.rpc_zmq_topic_backlog)
self.sockets.append(out_sock)
# It takes some time for a pub socket to open,
# before we can have any faith in doing a send() to it.
if sock_type == zmq.PUB:
eventlet.sleep(.5)
waiter.send(True)
while(True):
data = self.topic_proxy[topic].get()
out_sock.send(data, copy=False)
def _wait_child(self):
try:
# Don't block if no child processes have exited
pid, status = os.waitpid(0, os.WNOHANG)
if not pid:
return None
except OSError as exc:
if exc.errno not in (errno.EINTR, errno.ECHILD):
raise
return None
if os.WIFSIGNALED(status):
sig = os.WTERMSIG(status)
LOG.info(_('Child %(pid)d killed by signal %(sig)d'),
dict(pid=pid, sig=sig))
else:
code = os.WEXITSTATUS(status)
LOG.info(_('Child %(pid)s exited with status %(code)d'),
dict(pid=pid, code=code))
if pid not in self.children:
LOG.warning(_('pid %d not in child list'), pid)
return None
wrap = self.children.pop(pid)
wrap.children.remove(pid)
return wrap
def multicall(conf, context, topic, msg, timeout, connection_pool):
"""Make a call that returns multiple times."""
# TODO(pekowski): Remove all these comments in Havana.
# For amqp_rpc_single_reply_queue = False,
# Can't use 'with' for multicall, as it returns an iterator
# that will continue to use the connection. When it's done,
# connection.close() will get called which will put it back into
# the pool
# For amqp_rpc_single_reply_queue = True,
# The 'with' statement is mandatory for closing the connection
LOG.debug(_('Making synchronous call on %s ...'), topic)
msg_id = uuid.uuid4().hex
msg.update({'_msg_id': msg_id})
LOG.debug(_('MSG_ID is %s') % (msg_id))
_add_unique_id(msg)
pack_context(msg, context)
# TODO(pekowski): Remove this flag and the code under the if clause
# in Havana.
if not conf.amqp_rpc_single_reply_queue:
conn = ConnectionContext(conf, connection_pool)
wait_msg = MulticallWaiter(conf, conn, timeout)
conn.declare_direct_consumer(msg_id, wait_msg)
conn.topic_send(topic, rpc_common.serialize_msg(msg), timeout)
else:
with _reply_proxy_create_sem:
if not connection_pool.reply_proxy:
connection_pool.reply_proxy = ReplyProxy(conf, connection_pool)
msg.update({'_reply_q': connection_pool.reply_proxy.get_reply_q()})
wait_msg = MulticallProxyWaiter(conf, msg_id, timeout, connection_pool)
with ConnectionContext(conf, connection_pool) as conn:
conn.topic_send(topic, rpc_common.serialize_msg(msg), timeout)
return wait_msg
def _parse_check(rule):
"""
Parse a single base check rule into an appropriate Check object.
"""
# Handle the special checks
if rule == '!':
return FalseCheck()
elif rule == '@':
return TrueCheck()
try:
kind, match = rule.split(':', 1)
except Exception:
LOG.exception(_("Failed to understand rule %(rule)s") % locals())
# If the rule is invalid, we'll fail closed
return FalseCheck()
# Find what implements the check
if kind in _checks:
return _checks[kind](kind, match)
elif None in _checks:
return _checks[None](kind, match)
else:
LOG.error(_("No handler for matches of kind %s") % kind)
return FalseCheck()
def ssh_execute(ssh,
cmd,
process_input=None,
addl_env=None,
check_exit_code=True):
LOG.debug(_('Running cmd (SSH): %s'), cmd)
if addl_env:
raise InvalidArgumentError(_('Environment not supported over SSH'))
if process_input:
# This is (probably) fixable if we need it...
raise InvalidArgumentError(_('process_input not supported over SSH'))
stdin_stream, stdout_stream, stderr_stream = ssh.exec_command(cmd)
channel = stdout_stream.channel
# NOTE(justinsb): This seems suspicious...
# ...other SSH clients have buffering issues with this approach
stdout = stdout_stream.read()
stderr = stderr_stream.read()
stdin_stream.close()
exit_status = channel.recv_exit_status()
# exit_status == -1 if no exit code was returned
if exit_status != -1:
LOG.debug(_('Result was %s') % exit_status)
if check_exit_code and exit_status != 0:
raise ProcessExecutionError(exit_code=exit_status,
stdout=stdout,
stderr=stderr,
cmd=cmd)
return (stdout, stderr)
def __init__(self,
stdout=None,
stderr=None,
exit_code=None,
cmd=None,
description=None):
self.exit_code = exit_code
self.stderr = stderr
self.stdout = stdout
self.cmd = cmd
self.description = description
if description is None:
description = _("Unexpected error while running command.")
if exit_code is None:
exit_code = '-'
message = _('%(description)s\n'
'Command: %(cmd)s\n'
'Exit code: %(exit_code)s\n'
'Stdout: %(stdout)r\n'
'Stderr: %(stderr)r') % {
'description': description,
'cmd': cmd,
'exit_code': exit_code,
'stdout': stdout,
'stderr': stderr
}
super(ProcessExecutionError, self).__init__(message)
def register(self, proxy, in_addr, zmq_type_in, out_addr=None,
zmq_type_out=None, in_bind=True, out_bind=True,
subscribe=None):
LOG.info(_("Registering reactor"))
if zmq_type_in not in (zmq.PULL, zmq.SUB):
raise RPCException("Bad input socktype")
# Items push in.
inq = ZmqSocket(in_addr, zmq_type_in, bind=in_bind,
subscribe=subscribe)
self.proxies[inq] = proxy
self.sockets.append(inq)
LOG.info(_("In reactor registered"))
if not out_addr:
return
if zmq_type_out not in (zmq.PUSH, zmq.PUB):
raise RPCException("Bad output socktype")
# Items push out.
outq = ZmqSocket(out_addr, zmq_type_out, bind=out_bind)
self.mapping[inq] = outq
self.mapping[outq] = inq
self.sockets.append(outq)
LOG.info(_("Out reactor registered"))
def consume_in_thread(self):
"""Runs the ZmqProxy service"""
ipc_dir = CONF.rpc_zmq_ipc_dir
consume_in = "tcp://%s:%s" % \
(CONF.rpc_zmq_bind_address,
CONF.rpc_zmq_port)
consumption_proxy = InternalContext(None)
if not os.path.isdir(ipc_dir):
try:
utils.execute('mkdir', '-p', ipc_dir, run_as_root=True)
utils.execute('chown', "%s:%s" % (os.getuid(), os.getgid()),
ipc_dir, run_as_root=True)
utils.execute('chmod', '750', ipc_dir, run_as_root=True)
except utils.ProcessExecutionError:
with excutils.save_and_reraise_exception():
LOG.error(_("Could not create IPC directory %s") %
(ipc_dir, ))
try:
self.register(consumption_proxy,
consume_in,
zmq.PULL,
out_bind=True)
except zmq.ZMQError:
with excutils.save_and_reraise_exception():
LOG.error(_("Could not create ZeroMQ receiver daemon. "
"Socket may already be in use."))
super(ZmqProxy, self).consume_in_thread()
def acquire(self):
basedir = os.path.dirname(self.fname)
if not os.path.exists(basedir):
fileutils.ensure_tree(basedir)
LOG.info(_('Created lock path: %s'), basedir)
self.lockfile = open(self.fname, 'w')
while True:
try:
# Using non-blocking locks since green threads are not
# patched to deal with blocking locking calls.
# Also upon reading the MSDN docs for locking(), it seems
# to have a laughable 10 attempts "blocking" mechanism.
self.trylock()
LOG.debug(_('Got file lock "%s"'), self.fname)
return True
except IOError as e:
if e.errno in (errno.EACCES, errno.EAGAIN):
# external locks synchronise things like iptables
# updates - give it some time to prevent busy spinning
time.sleep(0.01)
else:
raise threading.ThreadError(_("Unable to acquire lock on"
" `%(filename)s` due to"
" %(exception)s") %
{
'filename': self.fname,
'exception': e,
})
def consume_in_thread(self):
"""Runs the ZmqProxy service"""
ipc_dir = CONF.rpc_zmq_ipc_dir
consume_in = "tcp://%s:%s" % \
(CONF.rpc_zmq_bind_address,
CONF.rpc_zmq_port)
consumption_proxy = InternalContext(None)
if not os.path.isdir(ipc_dir):
try:
utils.execute('mkdir', '-p', ipc_dir, run_as_root=True)
utils.execute('chown',
"%s:%s" % (os.getuid(), os.getgid()),
ipc_dir,
run_as_root=True)
utils.execute('chmod', '750', ipc_dir, run_as_root=True)
except utils.ProcessExecutionError:
with excutils.save_and_reraise_exception():
LOG.error(
_("Could not create IPC directory %s") % (ipc_dir, ))
try:
self.register(consumption_proxy,
consume_in,
zmq.PULL,
out_bind=True)
except zmq.ZMQError:
with excutils.save_and_reraise_exception():
LOG.error(
_("Could not create ZeroMQ receiver daemon. "
"Socket may already be in use."))
super(ZmqProxy, self).consume_in_thread()
def acquire(self):
basedir = os.path.dirname(self.fname)
if not os.path.exists(basedir):
fileutils.ensure_tree(basedir)
LOG.info(_('Created lock path: %s'), basedir)
self.lockfile = open(self.fname, 'w')
while True:
try:
# Using non-blocking locks since green threads are not
# patched to deal with blocking locking calls.
# Also upon reading the MSDN docs for locking(), it seems
# to have a laughable 10 attempts "blocking" mechanism.
self.trylock()
LOG.debug(_('Got file lock "%s"'), self.fname)
return True
except IOError as e:
if e.errno in (errno.EACCES, errno.EAGAIN):
# external locks synchronise things like iptables
# updates - give it some time to prevent busy spinning
time.sleep(0.01)
else:
raise threading.ThreadError(
_("Unable to acquire lock on"
" `%(filename)s` due to"
" %(exception)s") % {
'filename': self.fname,
'exception': e,
})
def __call__(self, message_data):
"""Consumer callback to call a method on a proxy object.
Parses the message for validity and fires off a thread to call the
proxy object method.
Message data should be a dictionary with two keys:
method: string representing the method to call
args: dictionary of arg: value
Example: {'method': 'echo', 'args': {'value': 42}}
"""
# It is important to clear the context here, because at this point
# the previous context is stored in local.store.context
if hasattr(local.store, 'context'):
del local.store.context
rpc_common._safe_log(LOG.debug, _('received %s'), message_data)
self.msg_id_cache.check_duplicate_message(message_data)
ctxt = unpack_context(self.conf, message_data)
method = message_data.get('method')
args = message_data.get('args', {})
version = message_data.get('version', None)
if not method:
LOG.warn(_('no method for message: %s') % message_data)
ctxt.reply(_('No method for message: %s') % message_data,
connection_pool=self.connection_pool)
return
self.pool.spawn_n(self._process_data, ctxt, version, method, args)
def check(rule, ctxt, target=None, do_raise=True, exc=exceptions.Forbidden):
creds = ctxt.to_dict()
target = target or {}
try:
result = _ENFORCER.enforce(rule, target, creds, do_raise, exc)
except Exception:
result = False
raise
else:
return result
finally:
extra = {'policy': {'rule': rule, 'target': target}}
if result:
LOG.audit(_("Policy check succeeded for rule '%(rule)s' "
"on target %(target)s") % {
'rule': rule,
'target': repr(target)
},
extra=extra)
else:
LOG.audit(_("Policy check failed for rule '%(rule)s' "
"on target %(target)s") % {
'rule': rule,
'target': repr(target)
},
extra=extra)
def _process_data(self, ctxt, version, method, args):
"""Process a message in a new thread.
If the proxy object we have has a dispatch method
(see rpc.dispatcher.RpcDispatcher), pass it the version,
method, and args and let it dispatch as appropriate. If not, use
the old behavior of magically calling the specified method on the
proxy we have here.
"""
ctxt.update_store()
try:
rval = self.proxy.dispatch(ctxt, version, method, **args)
# Check if the result was a generator
if inspect.isgenerator(rval):
for x in rval:
ctxt.reply(x, None, connection_pool=self.connection_pool)
else:
ctxt.reply(rval, None, connection_pool=self.connection_pool)
# This final None tells multicall that it is done.
ctxt.reply(ending=True, connection_pool=self.connection_pool)
except rpc_common.ClientException as e:
LOG.debug(
_('Expected exception during message handling (%s)') %
e._exc_info[1])
ctxt.reply(None,
e._exc_info,
connection_pool=self.connection_pool,
log_failure=False)
except Exception:
# sys.exc_info() is deleted by LOG.exception().
exc_info = sys.exc_info()
LOG.error(_('Exception during message handling'),
exc_info=exc_info)
ctxt.reply(None, exc_info, connection_pool=self.connection_pool)
def _wait_for_exit_or_signal(self, ready_callback=None):
status = None
signo = 0
LOG.debug(_('Full set of CONF:'))
CONF.log_opt_values(LOG, std_logging.DEBUG)
try:
if ready_callback:
ready_callback()
super(ServiceLauncher, self).wait()
except SignalExit as exc:
signame = _signo_to_signame(exc.signo)
LOG.info(_('Caught %s, exiting'), signame)
status = exc.code
signo = exc.signo
except SystemExit as exc:
status = exc.code
finally:
self.stop()
if rpc:
try:
rpc.cleanup()
except Exception:
# We're shutting down, so it doesn't matter at this point.
LOG.exception(_('Exception during rpc cleanup.'))
return status, signo
def multicall(conf, context, topic, msg, timeout, connection_pool):
"""Make a call that returns multiple times."""
# TODO(pekowski): Remove all these comments in Havana.
# For amqp_rpc_single_reply_queue = False,
# Can't use 'with' for multicall, as it returns an iterator
# that will continue to use the connection. When it's done,
# connection.close() will get called which will put it back into
# the pool
# For amqp_rpc_single_reply_queue = True,
# The 'with' statement is mandatory for closing the connection
LOG.debug(_('Making synchronous call on %s ...'), topic)
msg_id = uuid.uuid4().hex
msg.update({'_msg_id': msg_id})
LOG.debug(_('MSG_ID is %s') % (msg_id))
_add_unique_id(msg)
pack_context(msg, context)
# TODO(pekowski): Remove this flag and the code under the if clause
# in Havana.
if not conf.amqp_rpc_single_reply_queue:
conn = ConnectionContext(conf, connection_pool)
wait_msg = MulticallWaiter(conf, conn, timeout)
conn.declare_direct_consumer(msg_id, wait_msg)
conn.topic_send(topic, rpc_common.serialize_msg(msg), timeout)
else:
with _reply_proxy_create_sem:
if not connection_pool.reply_proxy:
connection_pool.reply_proxy = ReplyProxy(conf, connection_pool)
msg.update({'_reply_q': connection_pool.reply_proxy.get_reply_q()})
wait_msg = MulticallProxyWaiter(conf, msg_id, timeout, connection_pool)
with ConnectionContext(conf, connection_pool) as conn:
conn.topic_send(topic, rpc_common.serialize_msg(msg), timeout)
return wait_msg
def _start_child(self, wrap):
if len(wrap.forktimes) > wrap.workers:
# Limit ourselves to one process a second (over the period of
# number of workers * 1 second). This will allow workers to
# start up quickly but ensure we don't fork off children that
# die instantly too quickly.
if time.time() - wrap.forktimes[0] < wrap.workers:
LOG.info(_('Forking too fast, sleeping'))
time.sleep(1)
wrap.forktimes.pop(0)
wrap.forktimes.append(time.time())
pid = os.fork()
if pid == 0:
launcher = self._child_process(wrap.service)
while True:
self._child_process_handle_signal()
status, signo = self._child_wait_for_exit_or_signal(launcher)
if not _is_sighup_and_daemon(signo):
break
launcher.restart()
os._exit(status)
LOG.info(_('Started child %d'), pid)
wrap.children.add(pid)
self.children[pid] = wrap
return pid
def _error_callback(exc):
if isinstance(exc, qpid_exceptions.Empty):
LOG.debug(
_('Timed out waiting for RPC response: %s') % str(exc))
raise rpc_common.Timeout()
else:
LOG.exception(
_('Failed to consume message from queue: %s') % str(exc))
def release(self):
try:
self.unlock()
self.lockfile.close()
LOG.debug(_('Released file lock "%s"'), self.fname)
except IOError:
LOG.exception(_("Could not release the acquired lock `%s`"),
self.fname)
def _error_callback(exc):
if isinstance(exc, qpid_exceptions.Empty):
LOG.debug(_('Timed out waiting for RPC response: %s') %
str(exc))
raise rpc_common.Timeout()
else:
LOG.exception(_('Failed to consume message from queue: %s') %
str(exc))
def release(self):
try:
self.unlock()
self.lockfile.close()
LOG.debug(_('Released file lock "%s"'), self.fname)
except IOError:
LOG.exception(_("Could not release the acquired lock `%s`"),
self.fname)
def wrapper(*args, **kwargs):
try:
path = lock_path or os.environ.get("DESIGNATE_LOCK_PATH")
lock = lockfile.FileLock(os.path.join(path, lock_prefix))
with lock:
LOG.debug(_('Got lock "%s"') % f.__name__)
return f(*args, **kwargs)
finally:
LOG.debug(_('Lock released "%s"') % f.__name__)
def inner(*args, **kwargs):
try:
with lock(name, lock_file_prefix, external):
LOG.debug(_('Got semaphore / lock "%(function)s"'),
{'function': f.__name__})
return f(*args, **kwargs)
finally:
LOG.debug(_('Semaphore / lock released "%(function)s"'),
{'function': f.__name__})
def inner(*args, **kwargs):
try:
with lock(name, lock_file_prefix, external, lock_path):
LOG.debug(_('Got semaphore / lock "%(function)s"'),
{'function': f.__name__})
return f(*args, **kwargs)
finally:
LOG.debug(_('Semaphore / lock released "%(function)s"'),
{'function': f.__name__})
def _error_callback(exc):
if isinstance(exc, socket.timeout):
LOG.debug(_('Timed out waiting for RPC response: %s') %
str(exc))
raise rpc_common.Timeout()
else:
LOG.exception(_('Failed to consume message from queue: %s') %
str(exc))
info['do_consume'] = True
def wrapper(*args, **kwargs):
try:
path = lock_path or os.environ.get("DESIGNATE_LOCK_PATH")
lock = lockfile.FileLock(os.path.join(path, lock_prefix))
with lock:
LOG.debug(_('Got lock "%s"') % f.__name__)
return f(*args, **kwargs)
finally:
LOG.debug(_('Lock released "%s"') % f.__name__)
def reconnect(self):
"""Handles reconnecting and re-establishing queues.
Will retry up to self.max_retries number of times.
self.max_retries = 0 means to retry forever.
Sleep between tries, starting at self.interval_start
seconds, backing off self.interval_stepping number of seconds
each attempt.
"""
attempt = 0
while True:
params = self.params_list[attempt % len(self.params_list)]
attempt += 1
try:
self._connect(params)
return
except (IOError, self.connection_errors) as e:
pass
except Exception, e:
# NOTE(comstud): Unfortunately it's possible for amqplib
# to return an error not covered by its transport
# connection_errors in the case of a timeout waiting for
# a protocol response. (See paste link in LP888621)
# So, we check all exceptions for 'timeout' in them
# and try to reconnect in this case.
if 'timeout' not in str(e):
raise
log_info = {}
log_info['err_str'] = str(e)
log_info['max_retries'] = self.max_retries
log_info.update(params)
if self.max_retries and attempt == self.max_retries:
LOG.error(
_('Unable to connect to AMQP server on '
'%(hostname)s:%(port)d after %(max_retries)d '
'tries: %(err_str)s') % log_info)
# NOTE(comstud): Copied from original code. There's
# really no better recourse because if this was a queue we
# need to consume on, we have no way to consume anymore.
sys.exit(1)
if attempt == 1:
sleep_time = self.interval_start or 1
elif attempt > 1:
sleep_time += self.interval_stepping
if self.interval_max:
sleep_time = min(sleep_time, self.interval_max)
log_info['sleep_time'] = sleep_time
LOG.error(
_('AMQP server on %(hostname)s:%(port)d is '
'unreachable: %(err_str)s. Trying again in '
'%(sleep_time)d seconds.') % log_info)
time.sleep(sleep_time)
def _process_data(self, message_data):
msg_id = message_data.pop('_msg_id', None)
waiter = self._call_waiters.get(msg_id)
if not waiter:
LOG.warn(_('No calling threads waiting for msg_id : %(msg_id)s'
', message : %(data)s'), {'msg_id': msg_id,
'data': message_data})
LOG.warn(_('_call_waiters: %s') % str(self._call_waiters))
else:
waiter.put(message_data)
def reconnect(self):
"""Handles reconnecting and re-establishing queues.
Will retry up to self.max_retries number of times.
self.max_retries = 0 means to retry forever.
Sleep between tries, starting at self.interval_start
seconds, backing off self.interval_stepping number of seconds
each attempt.
"""
attempt = 0
while True:
params = self.params_list[attempt % len(self.params_list)]
attempt += 1
try:
self._connect(params)
return
except (IOError, self.connection_errors) as e:
pass
except Exception, e:
# NOTE(comstud): Unfortunately it's possible for amqplib
# to return an error not covered by its transport
# connection_errors in the case of a timeout waiting for
# a protocol response. (See paste link in LP888621)
# So, we check all exceptions for 'timeout' in them
# and try to reconnect in this case.
if 'timeout' not in str(e):
raise
log_info = {}
log_info['err_str'] = str(e)
log_info['max_retries'] = self.max_retries
log_info.update(params)
if self.max_retries and attempt == self.max_retries:
LOG.error(_('Unable to connect to AMQP server on '
'%(hostname)s:%(port)d after %(max_retries)d '
'tries: %(err_str)s') % log_info)
# NOTE(comstud): Copied from original code. There's
# really no better recourse because if this was a queue we
# need to consume on, we have no way to consume anymore.
sys.exit(1)
if attempt == 1:
sleep_time = self.interval_start or 1
elif attempt > 1:
sleep_time += self.interval_stepping
if self.interval_max:
sleep_time = min(sleep_time, self.interval_max)
log_info['sleep_time'] = sleep_time
LOG.error(_('AMQP server on %(hostname)s:%(port)d is '
'unreachable: %(err_str)s. Trying again in '
'%(sleep_time)d seconds.') % log_info)
time.sleep(sleep_time)
