version 0.5.1

• Fix a bug in the processes created by add_local_worker

  The _LocalProcess internal class was missing the expected process_callback attribute. Because the user has full control over the local process, the process_callback is unnecessary so the attribute was simply set to None.

version 0.5.0

• Adds support for local workers for any processing that is not driven by an AMQP queue.

def log_frob_rate(state):
    while True:
        logger.debug('Frobs: {0}/min'.format(state.get_frobs()))
        state.reset_frobs()
        time.sleep(60.0)

app.add_local_worker(log_frob_rate, args=[my_state])

• You can now override the name of the result queue with the result_queue argument to apply_async().

result = app.tasks.my_task.apply_async(['arg1', 'arg2'],
                                       send_results=True,
                                       result_queue='my_results')

• Task result queues are no longer implicitly deleted once the first result is retrieved with calls to AsyncResult methods. You should explicitly call result.delete() to delete the queue if needed.

result = app.tasks.my_task.apply_async(['arg1', 'arg2'], send_results=True)
print result.get()
result.delete()

• Task result queues may get results from many workers. This is useful for tasks that are published to many workers, where each worker will publish a result. The result.gather() method will yield new AsyncResult objects every time a result comes in.

results = app.tasks.my_task.apply_async(['arg1', 'arg2'],
                                        send_results=True,
                                        result_queue='my_results')
for res in results.gather():
    print res.get()

version 0.4.1

• Result queue prefix

  The result queue had been fixed with names like result_53fb9628-0760-44b9-b0f1-3539738f65b5. With this release, the result_ prefix is a class attribute and may be changed.

version 0.4.0

• Plugin reloading

  Re-loading the master process with a SIGHUP will now also reload all plugin modules.

• Renamed executable

  The provoke-worker process is know simply provoke. The arguments it takes are unchanged since version 0.3.2.

• Removed taskgroup concept

  The confusing taskgroup concept has been changed into simply an AMQP exchange and routing key pair. If neither are overridden, the default behavior delivers to the default exchange with the name of the task as the routing key.

• AMQP heartbeat thread

  In worker processes, a separate thread is responsible for sending heartbeats at the agreed upon interval.

version 0.3.2

• Changed to plugin system

  Previously, provoke attempted to import a Python module given on the command-line or in a config file. Now, using Plugin Discovery, known plugins are made available by name.

• Configurable logging

  A separate logging configuration file can be specified in a provoke configuration file, which will be loaded using logging.config.fileConfig.

version 0.2.3

Expanded Python support:

• CPython 2.6+
• CPython 3.3+

version 0.2.1

version 0.2.1

version 0.1.0

version 0.1.0

version 0.0.0

shuffle around where things happen