def _publish_and_retry_on_missing_exchange(self,
exchange,
msg,
routing_key=None,
timeout=None):
"""Publisher that retry if the exchange is missing.
"""
if not exchange.passive:
raise RuntimeError("_publish_and_retry_on_missing_exchange() must "
"be called with an passive exchange.")
# TODO(sileht): use @retrying
# NOTE(sileht): no need to wait the application expect a response
# before timeout is exshauted
duration = (timeout
if timeout is not None else self.kombu_reconnect_timeout)
timer = rpc_common.DecayingTimer(duration=duration)
timer.start()
while True:
try:
self._publish(exchange,
msg,
routing_key=routing_key,
timeout=timeout)
return
except self.connection.channel_errors as exc:
# NOTE(noelbk/sileht):
# If rabbit dies, the consumer can be disconnected before the
# publisher sends, and if the consumer hasn't declared the
# queue, the publisher's will send a message to an exchange
# that's not bound to a queue, and the message wll be lost.
# So we set passive=True to the publisher exchange and catch
# the 404 kombu ChannelError and retry until the exchange
# appears
if exc.code == 404 and timer.check_return() > 0:
LOG.info(
_LI("The exchange %(exchange)s to send to "
"%(routing_key)s doesn't exist yet, "
"retrying...") % {
'exchange': exchange.name,
'routing_key': routing_key
})
time.sleep(0.25)
continue
elif exc.code == 404:
msg = _("The exchange %(exchange)s to send to "
"%(routing_key)s still doesn't exist after "
"%(duration)s sec abandoning...") % {
'duration': duration,
'exchange': exchange.name,
'routing_key': routing_key
}
LOG.info(msg)
raise rpc_amqp.AMQPDestinationNotFound(msg)
raise
def test_duration_callback_with_args(self, now):
now.return_value = 0
t = common.DecayingTimer(2)
t.start()
now.return_value = 3
callback = mock.Mock()
remaining = t.check_return(callback, 1, a='b')
self.assertEqual(0, remaining)
callback.assert_called_once_with(1, a='b')
def iterconsume(self, limit=None, timeout=None):
"""Return an iterator that will consume from all queues/consumers.
NOTE(sileht): Must be called within the connection lock
"""
timer = rpc_common.DecayingTimer(duration=timeout)
timer.start()
def _raise_timeout(exc):
LOG.debug('Timed out waiting for RPC response: %s', exc)
raise rpc_common.Timeout()
def _recoverable_error_callback(exc):
self.do_consume = True
timer.check_return(_raise_timeout, exc)
def _error_callback(exc):
_recoverable_error_callback(exc)
LOG.error(('Failed to consume message from queue: %s'),
exc)
def _consume():
if self.do_consume:
queues_head = self.consumers[:-1] # not fanout.
queues_tail = self.consumers[-1] # fanout
for queue in queues_head:
queue.consume(nowait=True)
queues_tail.consume(nowait=False)
self.do_consume = False
poll_timeout = (self._poll_timeout if timeout is None
else min(timeout, self._poll_timeout))
while True:
if self._consume_loop_stopped:
self._consume_loop_stopped = False
raise StopIteration
if self._heartbeat_supported_and_enabled():
self._heartbeat_check()
try:
return self.connection.drain_events(timeout=poll_timeout)
except socket.timeout as exc:
poll_timeout = timer.check_return(
_raise_timeout, exc, maximum=self._poll_timeout)
for iteration in itertools.count(0):
if limit and iteration >= limit:
raise StopIteration
yield self.ensure(
_consume,
recoverable_error_callback=_recoverable_error_callback,
error_callback=_error_callback)
def consume(self, timeout=None):
"""Consume from all queues/consumers."""
timer = rpc_common.DecayingTimer(duration=timeout)
timer.start()
def _raise_timeout(exc):
LOG.debug('Timed out waiting for RPC response: %s', exc)
raise rpc_common.Timeout()
def _recoverable_error_callback(exc):
if not isinstance(exc, rpc_common.Timeout):
self._new_consumers = self._consumers
timer.check_return(_raise_timeout, exc)
def _error_callback(exc):
_recoverable_error_callback(exc)
LOG.error(_LE('Failed to consume message from queue: %s'),
exc)
def _consume():
# NOTE(sileht): in case the acknowledgment or requeue of a
# message fail, the kombu transport can be disconnected
# In this case, we must redeclare our consumers, so raise
# a recoverable error to trigger the reconnection code.
if not self.connection.connected:
raise self.connection.recoverable_connection_errors[0]
if self._new_consumers:
for tag, consumer in enumerate(self._consumers):
if consumer in self._new_consumers:
consumer.consume(tag=tag)
self._new_consumers = []
poll_timeout = (self._poll_timeout if timeout is None
else min(timeout, self._poll_timeout))
while True:
if self._consume_loop_stopped:
return
if self._heartbeat_supported_and_enabled():
self._heartbeat_check()
try:
self.connection.drain_events(timeout=poll_timeout)
return
except socket.timeout as exc:
poll_timeout = timer.check_return(
_raise_timeout, exc, maximum=self._poll_timeout)
with self._connection_lock:
self.ensure(_consume,
recoverable_error_callback=_recoverable_error_callback,
error_callback=_error_callback)
def test_reset(self, now):
now.return_value = 0
t = common.DecayingTimer(3)
t.start()
now.return_value = 1
remaining = t.check_return()
self.assertEqual(2, remaining)
t.restart()
remaining = t.check_return()
self.assertEqual(3, remaining)
def wait(self, msg_id, timeout, call_monitor_timeout):
# NOTE(sileht): for each msg_id we receive two amqp message
# first one with the payload, a second one to ensure the other
# have finish to send the payload
# NOTE(viktors): We are going to remove this behavior in the N
# release, but we need to keep backward compatibility, so we should
# support both cases for now.
timer = rpc_common.DecayingTimer(duration=timeout)
timer.start()
if call_monitor_timeout:
call_monitor_timer = rpc_common.DecayingTimer(
duration=call_monitor_timeout)
call_monitor_timer.start()
else:
call_monitor_timer = None
final_reply = None
ending = False
while not ending:
timeout = timer.check_return(self._raise_timeout_exception, msg_id)
if call_monitor_timer and timeout > 0:
cm_timeout = call_monitor_timer.check_return(
self._raise_timeout_exception, msg_id)
if cm_timeout < timeout:
timeout = cm_timeout
try:
message = self.waiters.get(msg_id, timeout=timeout)
except moves.queue.Empty:
self._raise_timeout_exception(msg_id)
reply, ending = self._process_reply(message)
if reply is not None:
# NOTE(viktors): This can be either first _send_reply() with an
# empty `result` field or a second _send_reply() with
# ending=True and no `result` field.
final_reply = reply
elif ending is False:
LOG.debug('Call monitor heartbeat received; '
'renewing timeout timer')
call_monitor_timer.restart()
return final_reply
def consume(self, timeout=None):
"""Consume from all queues/consumers."""
timer = rpc_common.DecayingTimer(duration=timeout)
timer.start()
def _raise_timeout(exc):
LOG.debug('Timed out waiting for RPC response: %s', exc)
raise rpc_common.Timeout()
def _recoverable_error_callback(exc):
self.do_consume = True
timer.check_return(_raise_timeout, exc)
def _error_callback(exc):
_recoverable_error_callback(exc)
LOG.error(_('Failed to consume message from queue: %s'), exc)
def _consume():
if self.do_consume:
queues_head = self.consumers[:-1] # not fanout.
queues_tail = self.consumers[-1] # fanout
for queue in queues_head:
queue.consume(nowait=True)
queues_tail.consume(nowait=False)
self.do_consume = False
poll_timeout = (self._poll_timeout if timeout is None else min(
timeout, self._poll_timeout))
while True:
if self._consume_loop_stopped:
return
if self._heartbeat_supported_and_enabled():
self.connection.heartbeat_check(
rate=self.driver_conf.heartbeat_rate)
try:
self.connection.drain_events(timeout=poll_timeout)
return
except socket.timeout as exc:
poll_timeout = timer.check_return(
_raise_timeout, exc, maximum=self._poll_timeout)
with self._connection_lock:
self.ensure(_consume,
recoverable_error_callback=_recoverable_error_callback,
error_callback=_error_callback)
def reply(self, reply=None, failure=None, log_failure=True):
if not self.msg_id:
# NOTE(Alexei_987) not sending reply, if msg_id is empty
# because reply should not be expected by caller side
return
# NOTE(sileht): return without hold the a connection if possible
if not self._obsolete_reply_queues.reply_q_valid(
self.reply_q, self.msg_id):
return
# NOTE(sileht): we read the configuration value from the driver
# to be able to backport this change in previous version that
# still have the qpid driver
duration = self.listener.driver.missing_destination_retry_timeout
timer = rpc_common.DecayingTimer(duration=duration)
timer.start()
while True:
try:
with self.listener.driver._get_connection(
rpc_common.PURPOSE_SEND) as conn:
self._send_reply(conn,
reply,
failure,
log_failure=log_failure)
return
except rpc_amqp.AMQPDestinationNotFound:
if timer.check_return() > 0:
LOG.debug(("The reply %(msg_id)s cannot be sent "
"%(reply_q)s reply queue don't exist, "
"retrying..."), {
'msg_id': self.msg_id,
'reply_q': self.reply_q
})
time.sleep(0.25)
else:
self._obsolete_reply_queues.add(self.reply_q, self.msg_id)
LOG.info(
_LI("The reply %(msg_id)s cannot be sent "
"%(reply_q)s reply queue don't exist after "
"%(duration)s sec abandoning..."), {
'msg_id': self.msg_id,
'reply_q': self.reply_q,
'duration': duration
})
return
def iterconsume(self, limit=None, timeout=None):
"""Return an iterator that will consume from all queues/consumers."""
timer = rpc_common.DecayingTimer(duration=timeout)
timer.start()
def _raise_timeout(exc):
LOG.debug('Timed out waiting for RPC response: %s', exc)
raise rpc_common.Timeout()
def _error_callback(exc):
timer.check_return(_raise_timeout, exc)
LOG.exception(_('Failed to consume message from queue: %s'), exc)
def _consume():
# NOTE(sileht):
# maximun value choosen according the best practice from kombu:
# http://kombu.readthedocs.org/en/latest/reference/kombu.common.html#kombu.common.eventloop
poll_timeout = 1 if timeout is None else min(timeout, 1)
while True:
if self._consume_loop_stopped:
self._consume_loop_stopped = False
raise StopIteration
try:
nxt_receiver = self.session.next_receiver(
timeout=poll_timeout)
except qpid_exceptions.Empty as exc:
poll_timeout = timer.check_return(_raise_timeout, exc,
maximum=1)
else:
break
try:
self._lookup_consumer(nxt_receiver).consume()
except Exception:
LOG.exception(_("Error processing message. "
"Skipping it."))
for iteration in itertools.count(0):
if limit and iteration >= limit:
raise StopIteration
yield self.ensure(_error_callback, _consume)
def wait(self, msg_id, timeout):
# NOTE(sileht): for each msg_id we receive two amqp message
# first one with the payload, a second one to ensure the other
# have finish to send the payload
timer = rpc_common.DecayingTimer(duration=timeout)
timer.start()
final_reply = None
ending = False
while not ending:
timeout = timer.check_return(self._raise_timeout_exception, msg_id)
try:
message = self.waiters.get(msg_id, timeout=timeout)
except moves.queue.Empty:
self._raise_timeout_exception(msg_id)
reply, ending = self._process_reply(message)
if not ending:
final_reply = reply
return final_reply
def consume(self, timeout=None):
"""recieve messages as many as max_fetch_messages.
In this functions, there are no while loop to subscribe.
This would be helpful when we wants to control the velocity of
subscription.
"""
duration = (self.consumer_timeout if timeout is None else timeout)
timer = driver_common.DecayingTimer(duration=duration)
timer.start()
def _raise_timeout():
LOG.debug('Timed out waiting for Kafka response')
raise driver_common.Timeout()
poll_timeout = (self.consumer_timeout if timeout is None else min(
timeout, self.consumer_timeout))
while True:
if self._consume_loop_stopped:
return
try:
next_timeout = poll_timeout * 1000.0
# TODO(use configure() method instead)
# Currently KafkaConsumer does not support for
# the case of updating only fetch_max_wait_ms parameter
self.consumer._config['fetch_max_wait_ms'] = next_timeout
messages = list(self.consumer.fetch_messages())
except Exception as e:
LOG.exception(_LE("Failed to consume messages: %s"), e)
messages = None
if not messages:
poll_timeout = timer.check_return(
_raise_timeout, maximum=self.consumer_timeout)
continue
return messages
def consume(self, timeout=None):
"""Receive up to 'max_fetch_messages' messages.
:param timeout: poll timeout in seconds
"""
duration = (self.consumer_timeout if timeout is None else timeout)
timer = driver_common.DecayingTimer(duration=duration)
timer.start()
def _raise_timeout():
LOG.debug('Timed out waiting for Kafka response')
raise driver_common.Timeout()
poll_timeout = (self.consumer_timeout if timeout is None
else min(timeout, self.consumer_timeout))
while True:
if self._consume_loop_stopped:
return
try:
next_timeout = poll_timeout * 1000.0
# TODO(use configure() method instead)
# Currently KafkaConsumer does not support for
# the case of updating only fetch_max_wait_ms parameter
self.consumer._config['fetch_max_wait_ms'] = next_timeout
messages = list(self.consumer.fetch_messages())
except Exception as e:
LOG.exception(_LE("Failed to consume messages: %s"), e)
messages = None
if not messages:
poll_timeout = timer.check_return(
_raise_timeout, maximum=self.consumer_timeout)
continue
return messages
def test_duration_is_none_and_maximun_set(self):
t = common.DecayingTimer()
t.start()
remaining = t.check_return(None, maximum=2)
self.assertEqual(2, remaining)
def test_duration_is_none(self):
t = common.DecayingTimer()
t.start()
remaining = t.check_return(None)
self.assertEqual(None, remaining)
def test_no_duration_but_maximum(self):
t = common.DecayingTimer()
t.start()
remaining = t.check_return(maximum=2)
self.assertEqual(2, remaining)
def test_no_duration_no_callback(self):
t = common.DecayingTimer()
t.start()
remaining = t.check_return()
self.assertEqual(None, remaining)
def reply(self, reply=None, failure=None):
if not self.msg_id:
# NOTE(Alexei_987) not sending reply, if msg_id is empty
# because reply should not be expected by caller side
return
# NOTE(sileht): return without hold the a connection if possible
if not self._obsolete_reply_queues.reply_q_valid(
self.reply_q, self.msg_id):
return
# NOTE(sileht): we read the configuration value from the driver
# to be able to backport this change in previous version that
# still have the qpid driver
duration = self.listener.driver.missing_destination_retry_timeout
timer = rpc_common.DecayingTimer(duration=duration)
timer.start()
while True:
try:
with self.listener.driver._get_connection(
rpc_common.PURPOSE_SEND, ) as conn:
self._send_reply(conn, reply, failure)
return
except oslo_messaging.MessageUndeliverable:
# queue not found
if timer.check_return() <= 0:
self._obsolete_reply_queues.add(self.reply_q, self.msg_id)
LOG.error(
'The reply %(msg_id)s failed to send after '
'%(duration)d seconds due to a missing queue '
'(%(reply_q)s). Abandoning...', {
'msg_id': self.msg_id,
'duration': duration,
'reply_q': self.reply_q
})
return
LOG.debug(
'The reply %(msg_id)s could not be sent due to a missing '
'queue (%(reply_q)s). Retrying...', {
'msg_id': self.msg_id,
'reply_q': self.reply_q
})
time.sleep(0.25)
except rpc_amqp.AMQPDestinationNotFound as exc:
# exchange not found/down
if timer.check_return() <= 0:
self._obsolete_reply_queues.add(self.reply_q, self.msg_id)
LOG.error(
'The reply %(msg_id)s failed to send after '
'%(duration)d seconds due to a broker issue '
'(%(exc)s). Abandoning...', {
'msg_id': self.msg_id,
'duration': duration,
'exc': exc
})
return
LOG.debug(
'The reply %(msg_id)s could not be sent due to a broker '
'issue (%(exc)s). Retrying...', {
'msg_id': self.msg_id,
'exc': exc
})
time.sleep(0.25)
def test_duration_expired_no_callback(self):
t = common.DecayingTimer(2)
t._ends_at = time.time() - 10
remaining = t.check_return()
self.assertAlmostEqual(-10, remaining, 0)
