I want process messages in few threads but i\'m getting error during execute this code:
from __future__ import with_statement
import pika
import sys
from pika.ad
This issue is generated because you have set the { noack: true }, but still trying to send acknowledgement.
For me, it was just that I told the queue I wasn't going to ack, then I acked.
E.g. WRONG:
channel.basic_consume(callback, queue=queue_name, no_ack=True)
and then in my callback:
def callback(ch, method, properties, body):
# do stuff
ch.basic_ack(delivery_tag = method.delivery_tag)
RIGHT:
channel.basic_consume(callback, queue=queue_name, no_ack=False)
Bottom line: If you want to manually ack, set no_ack=False.
From the docs:
no_ack: (bool) if set to True, automatic acknowledgement mode will be used (see http://www.rabbitmq.com/confirms.html)
After seeing RabbitMQ - upgraded to a new version and got a lot of "PRECONDITION_FAILED unknown delivery tag 1"
I changed my basic-consume to look like this:
consumer_tag = channel.basic_consume(
message_delivery_event,
no_ack=True,
queue=queue,
)
This had the effect of causing the described error on initial (not redelivered) acknowledgements when the message's delivery tag was specified. The delivery was extracted from the message delivery's method structure.
Using
channel.basic_ack(delivery_tag=0)
suppresses the error in this case, too
Looking at http://lists.rabbitmq.com/pipermail/rabbitmq-discuss/2011-July/013664.html makes it seem as though it may be an issue in RabbitMQ.
The problem probably is that you're setting no_ack=True
like this:
consumer_tag = channel.basic_consume(
message_delivery_event,
no_ack=True,
queue=queue,
)
And then acknowledging the messages:
channel.basic_ack(delivery_tag=args.delivery_tag)
You have to chose if you want to acknowledge or not and set the correct consume parameter.
There is a bug with your code. You share a channel across threads. This is not supported by pika (see FAQ). You have 2 options:
no_ack=True
flag in basic_get(...)
and do not use the channel object in thread's function doWork(...)
If you need to ACK message only after you have finished your work, then let the main thread (the while True:
loop) handle the message ack (and not the worker thread). Below is a modified version of your code that does that.
from __future__ import with_statement
import pika
import sys
from pika.adapters.blocking_connection import BlockingConnection
from pika import connection, credentials
import time
import threading
import random
from pika.adapters.select_connection import SelectConnection
from pika.connection import Connection
import traceback
from Queue import Queue, Empty
def doWork(body, args, channel, ack_queue):
time.sleep(random.random())
ack_queue.put(args.delivery_tag)
def doAck(channel):
while True:
try:
r = ack_queue.get_nowait()
except Empty:
r = None
if r is None:
break
try:
channel.basic_ack(delivery_tag=r)
except:
traceback.print_exc()
auth = credentials.PlainCredentials(username="guest", password="guest")
params = connection.ConnectionParameters(host="localhost", credentials=auth)
conn = BlockingConnection(params)
channel = conn.channel()
# Create a queue for the messages that should be ACKed by main thread
ack_queue = Queue()
while True:
time.sleep(0.03)
try:
doAck(channel)
method_frame, header_frame, body = channel.basic_get(queue="test_queue")
if method_frame.NAME == 'Basic.GetEmpty':
continue
t = threading.Thread(target=doWork, args=[body, method_frame, channel, ack_queue])
t.setDaemon(True)
t.start()
except Exception, e:
traceback.print_exc()
continue
You might also be encountering this error if you are trying to acknowledge a message on a different channel from which it was created. This might happen if you are closing or recreating channels.
From the docs: https://www.rabbitmq.com/confirms.html
Another scenario in which the broker will complain about an "unknown delivery tag" is when an acknowledgement, whether positive or negative, is attempted on a channel different from that on which the delivery was received on. Deliveries must be acknowledged on the same channel.