distributed.asyncio.AioClient seems to be broken

I am trying to execute distributed.asyncio.AioClient() inside a notebook and expect a future will be returned. However, it appears to be stuck forever.

When I interrupt the kernel I get the following traceback:

---------------------------------------------------------------------------
KeyboardInterrupt                         Traceback (most recent call last)
<ipython-input-2-544e7a0bbcaa> in <module>()
      1 import distributed.asyncio
----> 2 distributed.asyncio.AioClient()

/opt/conda/lib/python3.6/site-packages/distributed/asyncio.py in __init__(self, *args, **kwargs)
     96         loop = asyncio.get_event_loop()
     97         ioloop = BaseAsyncIOLoop(loop)
---> 98         super().__init__(*args, loop=ioloop, asynchronous=True, **kwargs)
     99 
    100     def __enter__(self):

/opt/conda/lib/python3.6/site-packages/distributed/client.py in __init__(self, address, loop, timeout, set_as_default, scheduler_file, security, asynchronous, name, **kwargs)
    553                                      io_loop=self.loop)
    554 
--> 555         self.start(timeout=timeout)
    556 
    557         from distributed.recreate_exceptions import ReplayExceptionClient

/opt/conda/lib/python3.6/site-packages/distributed/client.py in start(self, **kwargs)
    668             self._started = self._start(**kwargs)
    669         else:
--> 670             sync(self.loop, self._start, **kwargs)
    671 
    672     def __await__(self):

/opt/conda/lib/python3.6/site-packages/distributed/utils.py in sync(loop, func, *args, **kwargs)
    250     else:
    251         while not e.is_set():
--> 252             e.wait(1000000)
    253     if error[0]:
    254         six.reraise(*error[0])

/opt/conda/lib/python3.6/threading.py in wait(self, timeout)
    549             signaled = self._flag
    550             if not signaled:
--> 551                 signaled = self._cond.wait(timeout)
    552             return signaled
    553 

/opt/conda/lib/python3.6/threading.py in wait(self, timeout)
    297             else:
    298                 if timeout > 0:
--> 299                     gotit = waiter.acquire(True, timeout)
    300                 else:
    301                     gotit = waiter.acquire(False)

KeyboardInterrupt: 

distributed.utils - ERROR - sync() called from thread of running loop
Traceback (most recent call last):
  File "/opt/conda/lib/python3.6/site-packages/distributed/utils.py", line 235, in f
    raise RuntimeError("sync() called from thread of running loop")
RuntimeError: sync() called from thread of running loop

This part sync(self.loop, self._start, **kwargs) seems like it should not be executed. Also self.asynchronous=True, which seems incorrect.

awaiting the AioClient() also doesn’t work for me.

I’ve tried it on a colleagues computer (@jbweston), our Jupyterhub, and on my MacOS machine, all with the same result.

With distributed version 1.20.2 and dask version 0.16.1.

@mrocklin asked me to ping @kszucs.

[mrocklin]

It’s also worth noting that when tornado 5.0 is released (beta is out now) the tornado IOLoop will be the same as the asyncio event loop, so the need for AioClient should go away at that time.

[kszucs]

See Matthew’s comment, I haven’t tried it though.