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:
LOG.error(_("Could not create IPC directory %s") %
(ipc_dir, ))
raise
try:
self.register(consumption_proxy,
consume_in,
zmq.PULL,
out_bind=True)
except zmq.ZMQError:
LOG.error(_("Could not create ZeroMQ receiver daemon. "
"Socket may already be in use."))
raise
super(ZmqProxy, self).consume_in_thread()
def create_consumer(self, topic, proxy, fanout=False):
# Only consume on the base topic name.
topic = topic.split('.', 1)[0]
LOG.info(_("Create Consumer for topic (%(topic)s)") % {'topic': topic})
# Subscription scenarios
if fanout:
subscribe = ('', fanout)[type(fanout) == str]
sock_type = zmq.SUB
topic = 'fanout~' + topic
else:
sock_type = zmq.PULL
subscribe = None
# 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)
def wait(self):
"""Loop waiting on children to die and respawning as necessary"""
while self.running:
wrap = self._wait_child()
if not wrap:
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 reconnect(self):
"""Handles reconnecting and re-establishing sessions and queues"""
if self.connection.opened():
try:
self.connection.close()
except qpid.messaging.exceptions.ConnectionError:
pass
attempt = 0
delay = 1
while True:
broker = self.brokers[attempt % len(self.brokers)]
attempt += 1
try:
self.connection_create(broker)
self.connection.open()
except qpid.messaging.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 reconnect(self):
"""Handles reconnecting and re-establishing sessions and queues"""
if self.connection.opened():
try:
self.connection.close()
except qpid_exceptions.ConnectionError:
pass
attempt = 0
delay = 1
while True:
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 _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:
LOG.exception(_('Exception during message handling'))
ctxt.reply(None,
sys.exc_info(),
connection_pool=self.connection_pool)
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 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:
LOG.error(_("Could not create IPC directory %s") % (ipc_dir, ))
raise
try:
self.register(consumption_proxy,
consume_in,
zmq.PULL,
out_bind=True)
except zmq.ZMQError:
LOG.error(
_("Could not create ZeroMQ receiver daemon. "
"Socket may already be in use."))
raise
super(ZmqProxy, self).consume_in_thread()
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:
LOG.exception(_('Exception during message handling'))
ctxt.reply(None, sys.exc_info(),
connection_pool=self.connection_pool)
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)
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 _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 wait(self):
"""Loop waiting on children to die and respawning as necessary"""
while self.running:
wrap = self._wait_child()
if not wrap:
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 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 __init__(self, addr, zmq_type, bind=True, subscribe=None):
self.sock = ZMQ_CTX.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 _error_callback(exc):
if isinstance(exc, socket.timeout):
LOG.exception(_("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 publisher(waiter):
LOG.info(_("Creating proxy for topic: %s"), topic)
try:
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)
LOG.debug(_("ROUTER RELAY-OUT SUCCEEDED %(data)s") %
{'data': data})
def publisher(waiter):
LOG.info(_("Creating proxy for topic: %s"), topic)
try:
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)
LOG.debug(
_("ROUTER RELAY-OUT SUCCEEDED %(data)s") %
{'data': data})
def run_periodic_tasks(self, context, raise_on_error=False):
"""Tasks to be run at a periodic interval."""
idle_for = DEFAULT_INTERVAL
for task_name, task in self._periodic_tasks:
full_task_name = '.'.join([self.__class__.__name__, task_name])
now = timeutils.utcnow()
spacing = self._periodic_spacing[task_name]
last_run = self._periodic_last_run[task_name]
# If a periodic task is _nearly_ due, then we'll run it early
if spacing is not None and last_run is not None:
due = last_run + datetime.timedelta(seconds=spacing)
if not timeutils.is_soon(due, 0.2):
idle_for = min(idle_for, timeutils.delta_seconds(now, due))
continue
if spacing is not None:
idle_for = min(idle_for, spacing)
LOG.debug(_("Running periodic task %(full_task_name)s"), locals())
self._periodic_last_run[task_name] = timeutils.utcnow()
try:
task(self, context)
except Exception as e:
if raise_on_error:
raise
LOG.exception(_("Error during %(full_task_name)s: %(e)s"),
locals())
time.sleep(0)
return idle_for
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)
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 register_opts(conf):
"""Registration of options for this driver."""
#NOTE(ewindisch): ZMQ_CTX and matchmaker
# are initialized here as this is as good
# an initialization method as any.
# We memoize through these globals
global ZMQ_CTX
global matchmaker
global CONF
if not CONF:
conf.register_opts(zmq_opts)
CONF = conf
# Don't re-set, if this method is called twice.
if not ZMQ_CTX:
ZMQ_CTX = zmq.Context(conf.rpc_zmq_contexts)
if not matchmaker:
# rpc_zmq_matchmaker should be set to a 'module.Class'
mm_path = conf.rpc_zmq_matchmaker.split('.')
mm_module = '.'.join(mm_path[:-1])
mm_class = mm_path[-1]
# Only initialize a class.
if mm_path[-1][0] not in string.ascii_uppercase:
LOG.error(_("Matchmaker could not be loaded.\n"
"rpc_zmq_matchmaker is not a class."))
raise RPCException(_("Error loading Matchmaker."))
mm_impl = importutils.import_module(mm_module)
mm_constructor = getattr(mm_impl, mm_class)
matchmaker = mm_constructor()
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 _multi_send(method, context, topic, msg, timeout=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 = 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, _topic, msg, timeout)
return
return method(_addr, context, _topic, _topic, msg, timeout)
def create_consumer(self, topic, proxy, fanout=False):
# Only consume on the base topic name.
topic = topic.split('.', 1)[0]
LOG.info(_("Create Consumer for topic (%(topic)s)") %
{'topic': topic})
# Subscription scenarios
if fanout:
subscribe = ('', fanout)[type(fanout) == str]
sock_type = zmq.SUB
topic = 'fanout~' + topic
else:
sock_type = zmq.PULL
subscribe = None
# 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)
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 _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 _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.close()
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 _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 run_periodic_tasks(self, context, raise_on_error=False):
"""Tasks to be run at a periodic interval."""
idle_for = DEFAULT_INTERVAL
for task_name, task in self._periodic_tasks:
full_task_name = '.'.join([self.__class__.__name__, task_name])
now = timeutils.utcnow()
spacing = self._periodic_spacing[task_name]
last_run = self._periodic_last_run[task_name]
# If a periodic task is _nearly_ due, then we'll run it early
if spacing is not None and last_run is not None:
due = last_run + datetime.timedelta(seconds=spacing)
if not timeutils.is_soon(due, 0.2):
idle_for = min(idle_for, timeutils.delta_seconds(now, due))
continue
if spacing is not None:
idle_for = min(idle_for, spacing)
LOG.debug(_("Running periodic task %(full_task_name)s"), locals())
self._periodic_last_run[task_name] = timeutils.utcnow()
try:
task(self, context)
except Exception as e:
if raise_on_error:
raise
LOG.exception(_("Error during %(full_task_name)s: %(e)s"),
locals())
time.sleep(0)
return idle_for
def _error_callback(exc):
if isinstance(exc, qpid.messaging.exceptions.Empty):
LOG.exception(_('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 _error_callback(exc):
if isinstance(exc, socket.timeout):
LOG.exception(
_('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 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 notify(context, publisher_id, event_type, priority, payload):
"""Sends a notification using the specified driver
:param publisher_id: the source worker_type.host of the message
:param event_type: the literal type of event (ex. Instance Creation)
:param priority: patterned after the enumeration of Python logging
levels in the set (DEBUG, WARN, INFO, ERROR, CRITICAL)
:param payload: A python dictionary of attributes
Outgoing message format includes the above parameters, and appends the
following:
message_id
a UUID representing the id for this notification
timestamp
the GMT timestamp the notification was sent at
The composite message will be constructed as a dictionary of the above
attributes, which will then be sent via the transport mechanism defined
by the driver.
Message example::
{'message_id': str(uuid.uuid4()),
'publisher_id': 'compute.host1',
'timestamp': timeutils.utcnow(),
'priority': 'WARN',
'event_type': 'compute.create_instance',
'payload': {'instance_id': 12, ... }}
"""
if priority not in log_levels:
raise BadPriorityException(
_('%s not in valid priorities') % priority)
# Ensure everything is JSON serializable.
payload = jsonutils.to_primitive(payload, convert_instances=True)
msg = dict(message_id=str(uuid.uuid4()),
publisher_id=publisher_id,
event_type=event_type,
priority=priority,
payload=payload,
timestamp=str(timeutils.utcnow()))
for driver in _get_drivers():
try:
driver.notify(context, msg)
except Exception as e:
LOG.exception(_("Problem '%(e)s' attempting to "
"send to notification system. "
"Payload=%(payload)s")
% dict(e=e, payload=payload))
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 notify(context, message):
"""Deprecated in Grizzly. Please use rpc_notifier instead."""
LOG.deprecated(
_("The rabbit_notifier is now deprecated."
" Please use rpc_notifier instead."))
rpc_notifier.notify(context, message)
class Timeout(RPCException):
"""Signifies that a timeout has occurred.
This exception is raised if the rpc_response_timeout is reached while
waiting for a response from the remote side.
"""
message = _("Timeout while waiting on RPC response.")
def start(self, initial_delay=None, periodic_interval_max=None):
self._running = True
done = event.Event()
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, e:
self.stop()
done.send(e.retvalue)
except Exception:
LOG.exception(_('in dynamic looping call'))
done.send_exception(*sys.exc_info())
return
def _callback(raw_message):
message = self.channel.message_to_python(raw_message)
try:
callback(message.payload)
message.ack()
except Exception:
LOG.exception(_("Failed to process message... skipping it."))
def _callback(raw_message):
message = self.channel.message_to_python(raw_message)
try:
callback(message.payload)
message.ack()
except Exception:
LOG.exception(_("Failed to process message... skipping it."))
def start(self, interval, initial_delay=None):
self._running = True
done = event.Event()
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, e:
self.stop()
done.send(e.retvalue)
except Exception:
LOG.exception(_('in fixed duration looping call'))
done.send_exception(*sys.exc_info())
return
def _safe_log(log_func, msg, msg_data):
"""Sanitizes the msg_data field before logging."""
SANITIZE = {'set_admin_password': [('args', 'new_pass')],
'run_instance': [('args', 'admin_password')],
'route_message': [('args', 'message', 'args', 'method_info',
'method_kwargs', 'password'),
('args', 'message', 'args', 'method_info',
'method_kwargs', 'admin_password')]}
has_method = 'method' in msg_data and msg_data['method'] in SANITIZE
has_context_token = '_context_auth_token' in msg_data
has_token = 'auth_token' in msg_data
if not any([has_method, has_context_token, has_token]):
return log_func(msg, msg_data)
msg_data = copy.deepcopy(msg_data)
if has_method:
for arg in SANITIZE.get(msg_data['method'], []):
try:
d = msg_data
for elem in arg[:-1]:
d = d[elem]
d[arg[-1]] = '<SANITIZED>'
except KeyError, e:
LOG.info(_('Failed to sanitize %(item)s. Key error %(err)s'),
{'item': arg,
'err': e})
def serialize_remote_exception(failure_info, log_failure=True):
"""Prepares exception data to be sent over rpc.
Failure_info should be a sys.exc_info() tuple.
"""
tb = traceback.format_exception(*failure_info)
failure = failure_info[1]
if log_failure:
LOG.error(_("Returning exception %s to caller"), unicode(failure))
LOG.error(tb)
kwargs = {}
if hasattr(failure, 'kwargs'):
kwargs = failure.kwargs
data = {
'class': str(failure.__class__.__name__),
'module': str(failure.__class__.__module__),
'message': unicode(failure),
'tb': tb,
'args': failure.args,
'kwargs': kwargs
}
json_data = jsonutils.dumps(data)
return json_data
def deprecated(self, msg, *args, **kwargs):
stdmsg = _("Deprecated: %s") % msg
if CONF.fatal_deprecations:
self.critical(stdmsg, *args, **kwargs)
raise DeprecatedConfig(msg=stdmsg)
else:
self.warn(stdmsg, *args, **kwargs)
def deprecated(self, msg, *args, **kwargs):
stdmsg = _("Deprecated: %s") % msg
if CONF.fatal_deprecations:
self.critical(stdmsg, *args, **kwargs)
raise DeprecatedConfig(msg=stdmsg)
else:
self.warn(stdmsg, *args, **kwargs)
def start(self, interval, initial_delay=None):
self._running = True
done = event.Event()
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, e:
self.stop()
done.send(e.retvalue)
except Exception:
LOG.exception(_('in fixed duration looping call'))
done.send_exception(*sys.exc_info())
return
def notify(conf, context, topic, msg, connection_pool):
"""Sends a notification event on a topic."""
event_type = msg.get("event_type")
LOG.debug(_("Sending %(event_type)s on %(topic)s"), locals())
pack_context(msg, context)
with ConnectionContext(conf, connection_pool) as conn:
conn.notify_send(topic, msg)
def notify(conf, context, topic, msg, connection_pool):
"""Sends a notification event on a topic."""
event_type = msg.get('event_type')
LOG.debug(_('Sending %(event_type)s on %(topic)s'), locals())
pack_context(msg, context)
with ConnectionContext(conf, connection_pool) as conn:
conn.notify_send(topic, msg)
def start(self, initial_delay=None, periodic_interval_max=None):
self._running = True
done = event.Event()
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, e:
self.stop()
done.send(e.retvalue)
except Exception:
LOG.exception(_('in dynamic looping call'))
done.send_exception(*sys.exc_info())
return
class InternalContext(object):
"""Used by ConsumerBase as a private context for - methods."""
def __init__(self, proxy):
self.proxy = proxy
self.msg_waiter = None
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['args'])
return ConsumerBase.normalize_reply(result, ctx.replies)
except greenlet.GreenletExit:
# ignore these since they are just from shutdowns
pass
except rpc_common.ClientException, 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 __init__(cls, names, bases, dict_):
"""Metaclass that allows us to collect decorated periodic tasks."""
super(_PeriodicTasksMeta, cls).__init__(names, bases, dict_)
# NOTE(sirp): if the attribute is not present then we must be the base
# class, so, go ahead an initialize it. If the attribute is present,
# then we're a subclass so make a copy of it so we don't step on our
# parent's toes.
try:
cls._periodic_tasks = cls._periodic_tasks[:]
except AttributeError:
cls._periodic_tasks = []
try:
cls._periodic_last_run = cls._periodic_last_run.copy()
except AttributeError:
cls._periodic_last_run = {}
try:
cls._periodic_spacing = cls._periodic_spacing.copy()
except AttributeError:
cls._periodic_spacing = {}
for value in cls.__dict__.values():
if getattr(value, '_periodic_task', False):
task = value
name = task.__name__
if task._periodic_spacing < 0:
LOG.info(
_('Skipping periodic task %(task)s because '
'its interval is negative'), {'task': name})
continue
if not task._periodic_enabled:
LOG.info(
_('Skipping periodic task %(task)s because '
'it is disabled'), {'task': name})
continue
# A periodic spacing of zero indicates that this task should
# be run every pass
if task._periodic_spacing == 0:
task._periodic_spacing = None
cls._periodic_tasks.append((name, task))
cls._periodic_spacing[name] = task._periodic_spacing
cls._periodic_last_run[name] = task._periodic_last_run
def __init__(cls, names, bases, dict_):
"""Metaclass that allows us to collect decorated periodic tasks."""
super(_PeriodicTasksMeta, cls).__init__(names, bases, dict_)
# NOTE(sirp): if the attribute is not present then we must be the base
# class, so, go ahead an initialize it. If the attribute is present,
# then we're a subclass so make a copy of it so we don't step on our
# parent's toes.
try:
cls._periodic_tasks = cls._periodic_tasks[:]
except AttributeError:
cls._periodic_tasks = []
try:
cls._periodic_last_run = cls._periodic_last_run.copy()
except AttributeError:
cls._periodic_last_run = {}
try:
cls._periodic_spacing = cls._periodic_spacing.copy()
except AttributeError:
cls._periodic_spacing = {}
for value in cls.__dict__.values():
if getattr(value, '_periodic_task', False):
task = value
name = task.__name__
if task._periodic_spacing < 0:
LOG.info(_('Skipping periodic task %(task)s because '
'its interval is negative'),
{'task': name})
continue
if not task._periodic_enabled:
LOG.info(_('Skipping periodic task %(task)s because '
'it is disabled'),
{'task': name})
continue
# A periodic spacing of zero indicates that this task should
# be run every pass
if task._periodic_spacing == 0:
task._periodic_spacing = None
cls._periodic_tasks.append((name, task))
cls._periodic_spacing[name] = task._periodic_spacing
cls._periodic_last_run[name] = task._periodic_last_run
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
msg_id, topic, style, in_msg = data
ctx, request = rpc_common.deserialize_msg(_deserialize(in_msg))
ctx = RpcContext.unmarshal(ctx)
proxy = self.proxies[sock]
self.pool.spawn_n(self.process, style, topic, proxy, ctx, request)
def _callback(raw_message):
message = self.channel.message_to_python(raw_message)
try:
msg = rpc_common.deserialize_msg(message.payload)
callback(msg)
message.ack()
except Exception:
LOG.exception(_("Failed to process message... skipping it."))
def multicall(conf, context, topic, msg, timeout, connection_pool):
"""Make a call that returns multiple times."""
# 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
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))
pack_context(msg, context)
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))
return wait_msg
def _pipe_watcher(self):
# This will block until the write end is closed when the parent
# dies unexpectedly
self.readpipe.read()
LOG.info(_('Parent process has died unexpectedly, exiting'))
sys.exit(1)
def multicall(conf, context, topic, msg, timeout, connection_pool):
"""Make a call that returns multiple times."""
# 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
LOG.debug(_('Making asynchronous call on %s ...'), topic)
msg_id = uuid.uuid4().hex
msg.update({'_msg_id': msg_id})
LOG.debug(_('MSG_ID is %s') % (msg_id))
pack_context(msg, context)
conn = ConnectionContext(conf, connection_pool)
wait_msg = MulticallWaiter(conf, conn, timeout)
conn.declare_direct_consumer(msg_id, wait_msg)
conn.topic_send(topic, msg)
return wait_msg
def _pipe_watcher(self):
# This will block until the write end is closed when the parent
# dies unexpectedly
self.readpipe.read()
LOG.info(_('Parent process has died unexpectedly, exiting'))
sys.exit(1)