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 __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 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 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 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 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(0.5)
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 _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 __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 _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 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 _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 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 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 _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 _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 _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 _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 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 as 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 __init__(self, info=None, topic=None, method=None):
"""
:param info: Extra info to convey to the user
:param topic: The topic that the rpc call was sent to
:param rpc_method_name: The name of the rpc method being
called
"""
self.info = info
self.topic = topic
self.method = method
super(Timeout, self).__init__(
None,
info=info or _('<unknown>'),
topic=topic or _('<unknown>'),
method=method or _('<unknown>'))
def _get_socket(self, host, port, backlog):
# TODO(dims): eventlet's green dns/socket module does not actually
# support IPv6 in getaddrinfo(). We need to get around this in the
# future or monitor upstream for a fix
info = socket.getaddrinfo(host, port, socket.AF_UNSPEC, socket.SOCK_STREAM)[0]
family = info[0]
bind_addr = info[-1]
sock = None
retry_until = time.time() + 30
while not sock and time.time() < retry_until:
try:
sock = eventlet.listen(bind_addr, backlog=backlog, family=family)
if sslutils.is_enabled():
sock = sslutils.wrap(sock)
except socket.error as err:
if err.args[0] != errno.EADDRINUSE:
raise
eventlet.sleep(0.1)
if not sock:
raise RuntimeError(
_("Could not bind to %(host)s:%(port)s " "after trying for 30 seconds") % {"host": host, "port": port}
)
sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
# sockets can hang around forever without keepalive
sock.setsockopt(socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1)
# This option isn't available in the OS X version of eventlet
if hasattr(socket, "TCP_KEEPIDLE"):
sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_KEEPIDLE, CONF.tcp_keepidle)
return sock
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"),
six.text_type(failure))
LOG.error(tb)
kwargs = {}
if hasattr(failure, 'kwargs'):
kwargs = failure.kwargs
data = {
'class': str(failure.__class__.__name__),
'module': str(failure.__class__.__module__),
'message': six.text_type(failure),
'tb': tb,
'args': failure.args,
'kwargs': kwargs
}
json_data = jsonutils.dumps(data)
return json_data
def run(self, key):
# Assume starts with "fanout~", strip it for lookup.
nkey = key.split("fanout~")[1:][0]
if not self._ring_has(nkey):
LOG.warn(_("No key defining hosts for topic '%s', " "see ringfile") % (nkey,))
return []
return map(lambda x: (key + "." + x, x), self.ring[nkey])
def fanout_cast(conf, context, topic, msg, connection_pool):
"""Sends a message on a fanout exchange without waiting for a response."""
LOG.debug(_('Making asynchronous fanout cast...'))
_add_unique_id(msg)
pack_context(msg, context)
with ConnectionContext(conf, connection_pool) as conn:
conn.fanout_send(topic, rpc_common.serialize_msg(msg))
def cast(conf, context, topic, msg, connection_pool):
"""Sends a message on a topic without waiting for a response."""
LOG.debug(_('Making asynchronous cast on %s...'), topic)
_add_unique_id(msg)
pack_context(msg, context)
with ConnectionContext(conf, connection_pool) as conn:
conn.topic_send(topic, rpc_common.serialize_msg(msg))
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 read(self, i=None):
result = self.data.read(i)
self.bytes_read += len(result)
if self.bytes_read > self.limit:
msg = _("Request is too large.")
raise webob.exc.HTTPRequestEntityTooLarge(explanation=msg)
return result
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 add_call_waiter(self, waiter, msg_id):
self._num_call_waiters += 1
if self._num_call_waiters > self._num_call_waiters_wrn_threshhold:
LOG.warn(_('Number of call waiters is greater than warning '
'threshhold: %d. There could be a MulticallProxyWaiter '
'leak.') % self._num_call_waiters_wrn_threshhold)
self._num_call_waiters_wrn_threshhold *= 2
self._call_waiters[msg_id] = waiter
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 : %s'
', message : %s') % (msg_id, message_data))
else:
waiter.put(message_data)
def __iter__(self):
for chunk in self.data:
self.bytes_read += len(chunk)
if self.bytes_read > self.limit:
msg = _("Request is too large.")
raise webob.exc.HTTPRequestEntityTooLarge(explanation=msg)
else:
yield chunk