consumer: handler blocks and connection is stale

For my Message handler need quit a long time to process, I enable the message asynchronously as follows:

def handler(message):
    message.enable_async()
    # process it for a long time
    deal_with_task()
    message.finish()

and my nsq reader set like this

nsq.Reader(
    message_handler=self.handler,
    lookupd_http_addresses=NSQ_LOOKUP_HOST,
    channel=LISTEN_CHANNEL,
    topic=LISTEN_TOPIC,
    lookupd_poll_interval=100,
    msg_timeout=3600,
    max_tries=3,
    max_in_flight=1,
)
nsq.run()

The program may get the error like this in some cases:

WARNING:nsq.client:[127.0.0.1:4152] connection is stale (298.42s), closing
ERROR:nsq.async:uncaught exception in data event
Traceback (most recent call last):
  File "/usr/local/lib/python2.7/site-packages/nsq/async.py", line 275, in _read_body
    self.trigger(event.DATA, conn=self, data=data)
  File "/usr/local/lib/python2.7/site-packages/nsq/event.py", line 85, in trigger
    ev(*args, **kwargs)
  File "/usr/local/lib/python2.7/site-packages/nsq/async.py", line 477, in _on_data
    self.send(protocol.nop())
  File "/usr/local/lib/python2.7/site-packages/nsq/async.py", line 281, in send
    self.stream.write(data)
  File "/usr/local/lib/python2.7/site-packages/tornado/iostream.py", line 377, in write
    self._check_closed()
  File "/usr/local/lib/python2.7/site-packages/tornado/iostream.py", line 885, in _check_closed
    raise StreamClosedError(real_error=self.error)
StreamClosedError: Stream is closed

I don’t know what’s wrong with the program. And I am sure that there is no error uncaught while processing the task. The error happens when nsq client is listening on the nsqd. If anyone needs more codes, I can paste here.

[alpaker]

So,if I want to use 121 seconds to do my job,and I set the msg-message=“120s”.The nsqd also force to requeue the message,does it mean I should set the msg-message>121 seconds?

Yes. The argument to pass to to nsq.Reader() is msg_timeout.

By the way,do i use the message.enable_async() in a correct way?

Yes, except for where you’re invoking finish(). In general it’s not safe to a call a method that interacts with a Tornado IO loop (as finish() does) from a thread other the thread that IO loop is running on (in this case, the thread the Reader is running on).

You don’t want to use Python threads directly, but rather via a tornado executor facilities. An example is here. You’ll want to invoke finish() in the callback you pass to run_background().

[patterson163]

@alpaker Thanks for reply.