def listen(addr, handle_comm, deserialize=True, **kwargs):
"""
Create a listener object with the given parameters. When its ``start()``
method is called, the listener will listen on the given address
(a URI such as ``tcp://0.0.0.0``) and call *handle_comm* with a
``Comm`` object for each incoming connection.
*handle_comm* can be a regular function or a coroutine.
"""
try:
scheme, loc = parse_address(addr, strict=True)
except ValueError:
if kwargs.get("ssl_context"):
addr = "tls://" + addr
else:
addr = "tcp://" + addr
scheme, loc = parse_address(addr, strict=True)
backend = registry.get_backend(scheme)
return backend.get_listener(loc, handle_comm, deserialize, **kwargs)
async def _create_listeners(session_state, nworkers, rank):
assert session_state["loop"] is asyncio.get_event_loop()
assert "nworkers" not in session_state
session_state["nworkers"] = nworkers
assert "rank" not in session_state
session_state["rank"] = rank
async def server_handler(ep):
peer_rank = await ep.read()
session_state["eps"][peer_rank] = ep
# We listen on the same protocol and address as the worker address
protocol, address = parse_address(session_state["worker"].address)
address = parse_host_port(address)[0]
address = unparse_address(protocol, address)
session_state["lf"] = distributed.comm.listen(address, server_handler)
await session_state["lf"].start()
return session_state["lf"].listen_address
def parse_it(x):
return parse_host_port(parse_address(x)[1])
async def connect(
addr, timeout=None, deserialize=True, handshake_overrides=None, **connection_args
):
"""
Connect to the given address (a URI such as ``tcp://127.0.0.1:1234``)
and yield a ``Comm`` object. If the connection attempt fails, it is
retried until the *timeout* is expired.
"""
if timeout is None:
timeout = dask.config.get("distributed.comm.timeouts.connect")
timeout = parse_timedelta(timeout, default="seconds")
scheme, loc = parse_address(addr)
backend = registry.get_backend(scheme)
connector = backend.get_connector()
comm = None
start = time()
def time_left():
deadline = start + timeout
return max(0, deadline - time())
backoff_base = 0.01
attempt = 0
# Prefer multiple small attempts than one long attempt. This should protect
# primarily from DNS race conditions
# gh3104, gh4176, gh4167
intermediate_cap = timeout / 5
active_exception = None
while time_left() > 0:
try:
comm = await asyncio.wait_for(
connector.connect(loc, deserialize=deserialize, **connection_args),
timeout=min(intermediate_cap, time_left()),
)
break
except FatalCommClosedError:
raise
# Note: CommClosed inherits from OSError
except (asyncio.TimeoutError, OSError) as exc:
active_exception = exc
# As descibed above, the intermediate timeout is used to distributed
# initial, bulk connect attempts homogeneously. In particular with
# the jitter upon retries we should not be worred about overloading
# any more DNS servers
intermediate_cap = timeout
# FullJitter see https://aws.amazon.com/blogs/architecture/exponential-backoff-and-jitter/
upper_cap = min(time_left(), backoff_base * (2**attempt))
backoff = random.uniform(0, upper_cap)
attempt += 1
logger.debug(
"Could not connect to %s, waiting for %s before retrying", loc, backoff
)
await asyncio.sleep(backoff)
else:
raise OSError(
f"Timed out trying to connect to {addr} after {timeout} s"
) from active_exception
local_info = {
**comm.handshake_info(),
**(handshake_overrides or {}),
}
try:
# This would be better, but connections leak if worker is closed quickly
# write, handshake = await asyncio.gather(comm.write(local_info), comm.read())
handshake = await asyncio.wait_for(comm.read(), time_left())
await asyncio.wait_for(comm.write(local_info), time_left())
except Exception as exc:
with suppress(Exception):
await comm.close()
raise OSError(
f"Timed out during handshake while connecting to {addr} after {timeout} s"
) from exc
comm.remote_info = handshake
comm.remote_info["address"] = comm._peer_addr
comm.local_info = local_info
comm.local_info["address"] = comm._local_addr
comm.handshake_options = comm.handshake_configuration(
comm.local_info, comm.remote_info
)
return comm