def filesystem_receiver(logdir: str, q: "queue.Queue[AddressedMonitoringMessage]", run_dir: str) -> None:
logger = start_file_logger("{}/monitoring_filesystem_radio.log".format(logdir),
name="monitoring_filesystem_radio",
level=logging.DEBUG)
logger.info("Starting filesystem radio receiver")
setproctitle("parsl: monitoring filesystem receiver")
base_path = f"{run_dir}/monitor-fs-radio/"
tmp_dir = f"{base_path}/tmp/"
new_dir = f"{base_path}/new/"
logger.debug(f"Creating new and tmp paths under {base_path}")
os.makedirs(tmp_dir, exist_ok=True)
os.makedirs(new_dir, exist_ok=True)
while True: # this loop will end on process termination
logger.info("Start filesystem radio receiver loop")
# iterate over files in new_dir
for filename in os.listdir(new_dir):
try:
logger.info(f"Processing filesystem radio file {filename}")
full_path_filename = f"{new_dir}/{filename}"
with open(full_path_filename, "rb") as f:
message = deserialize(f.read())
logger.info(f"Message received is: {message}")
assert(isinstance(message, tuple))
q.put(cast(AddressedMonitoringMessage, message))
os.remove(full_path_filename)
except Exception:
logger.exception(f"Exception processing {filename} - probably will be retried next iteration")
time.sleep(1) # whats a good time for this poll?
def _queue_management_worker(self):
""" TODO: docstring """
logger.debug("[MTHREAD] queue management worker starting")
while not self.bad_state_is_set:
task_id, buf = self.incoming_q.get() # TODO: why does this hang?
msg = deserialize(buf)[0]
# TODO: handle exceptions
task_fut = self.tasks[task_id]
logger.debug("Got response for task id {}".format(task_id))
if "result" in msg:
task_fut.set_result(msg["result"])
elif "exception" in msg:
# TODO: handle exception
pass
elif 'exception' in msg:
logger.warning("Task: {} has returned with an exception")
try:
s = deserialize(msg['exception'])
exception = ValueError(
"Remote exception description: {}".format(s))
task_fut.set_exception(exception)
except Exception as e:
# TODO could be a proper wrapped exception?
task_fut.set_exception(
DeserializationError(
"Received exception, but handling also threw an exception: {}"
.format(e)))
else:
raise BadMessage(
"Message received is neither result nor exception")
if not self.is_alive:
break
logger.info("[MTHREAD] queue management worker finished")
def _complete_future(
expected_file: str, future_wrapper: FluxFutureWrapper, flux_future: Any
):
"""Callback triggered when a FluxExecutorFuture completes.
When the FluxExecutorFuture completes, check for the Parsl task's
output file, and assign the result to the FluxWrapperFuture future.
Parameters
----------
expected_file : str
The path to the Parsl task's output file, storing the result of the task.
future_wrapper : FluxFutureWrapper
The user-facing future.
flux_future : FluxExecutorFuture
The future wrapped by ``future_wrapper``. Also accessible via
``future_wrapper``, but the flux_future must be accepted as an argument
due to how ``concurrent.futures.add_done_callback`` works.
"""
if flux_future.cancelled(): # if underlying future was cancelled, return
return # no need to set a result on the wrapper future
try:
returncode = flux_future.result()
except Exception as unknown_err:
future_wrapper.set_exception(unknown_err)
return
if returncode == 0:
try: # look for the output file
with open(expected_file, "rb") as file_handle:
task_result = deserialize(file_handle.read())
except FileNotFoundError:
future_wrapper.set_exception(
FileNotFoundError(
f"No result found for Parsl task, expected {expected_file}"
)
)
except Exception as unknown_err:
future_wrapper.set_exception(unknown_err)
else: # task package deserialized successfully
if task_result.exception is not None:
future_wrapper.set_exception(task_result.exception)
else:
future_wrapper.set_result(task_result.returnval)
else: # the job exited abnormally
future_wrapper.set_exception(
AppException(f"Parsl task exited abnormally: returned {returncode}")
)
def runner(incoming_q, outgoing_q):
"""This is a function that mocks the Swift-T side.
It listens on the the incoming_q for tasks and posts returns on the outgoing_q.
Args:
- incoming_q (Queue object) : The queue to listen on
- outgoing_q (Queue object) : Queue to post results on
The messages posted on the incoming_q will be of the form :
.. code:: python
{
"task_id" : <uuid.uuid4 string>,
"buffer" : serialized buffer containing the fn, args and kwargs
}
If ``None`` is received, the runner will exit.
Response messages should be of the form:
.. code:: python
{
"task_id" : <uuid.uuid4 string>,
"result" : serialized buffer containing result
"exception" : serialized exception object
}
On exiting the runner will post ``None`` to the outgoing_q
"""
logger.debug("[RUNNER] Starting")
def execute_task(bufs):
"""Deserialize the buffer and execute the task.
Returns the serialized result or exception.
"""
user_ns = locals()
user_ns.update({'__builtins__': __builtins__})
f, args, kwargs = unpack_apply_message(bufs, user_ns, copy=False)
fname = getattr(f, '__name__', 'f')
prefix = "parsl_"
fname = prefix + "f"
argname = prefix + "args"
kwargname = prefix + "kwargs"
resultname = prefix + "result"
user_ns.update({
fname: f,
argname: args,
kwargname: kwargs,
resultname: resultname
})
code = "{0} = {1}(*{2}, **{3})".format(resultname, fname, argname,
kwargname)
try:
logger.debug("[RUNNER] Executing: {0}".format(code))
exec(code, user_ns, user_ns)
except Exception as e:
logger.warning("Caught exception; will raise it: {}".format(e))
raise e
else:
logger.debug("[RUNNER] Result: {0}".format(
user_ns.get(resultname)))
return user_ns.get(resultname)
while True:
try:
# Blocking wait on the queue
msg = incoming_q.get(block=True, timeout=10)
except queue.Empty:
# Handle case where no items were in the queue
logger.debug("[RUNNER] Queue is empty")
except IOError as e:
logger.debug("[RUNNER] Broken pipe: {}".format(e))
try:
# Attempt to send a stop notification to the management thread
outgoing_q.put(None)
except Exception:
pass
break
except Exception as e:
logger.debug("[RUNNER] Caught unknown exception: {}".format(e))
else:
# Handle received message
if not msg:
# Empty message is a die request
logger.debug("[RUNNER] Received exit request")
outgoing_q.put(None)
break
else:
# Received a valid message, handle it
logger.debug("[RUNNER] Got a valid task with ID {}".format(
msg["task_id"]))
try:
response_obj = execute_task(msg['buffer'])
response = {
"task_id": msg["task_id"],
"result": serialize(response_obj)
}
logger.debug("[RUNNER] Returing result: {}".format(
deserialize(response["result"])))
except Exception as e:
logger.debug(
"[RUNNER] Caught task exception: {}".format(e))
response = {
"task_id": msg["task_id"],
"exception": serialize(e)
}
outgoing_q.put(response)
logger.debug("[RUNNER] Terminating")
def _queue_management_worker(self):
"""Listen to the queue for task status messages and handle them.
Depending on the message, tasks will be updated with results, exceptions,
or updates. It expects the following messages:
.. code:: python
{
"task_id" : <task_id>
"result" : serialized result object, if task succeeded
... more tags could be added later
}
{
"task_id" : <task_id>
"exception" : serialized exception object, on failure
}
We do not support these yet, but they could be added easily.
.. code:: python
{
"task_id" : <task_id>
"cpu_stat" : <>
"mem_stat" : <>
"io_stat" : <>
"started" : tstamp
}
The `None` message is a die request.
"""
while True:
logger.debug("[MTHREAD] Management thread active")
try:
msg = self.incoming_q.get(block=True, timeout=1)
except queue.Empty:
# Timed out.
pass
except IOError as e:
logger.debug(
"[MTHREAD] Caught broken queue with exception code {}: {}".
format(e.errno, e))
return
except Exception as e:
logger.debug(
"[MTHREAD] Caught unknown exception: {}".format(e))
else:
if msg is None:
logger.debug("[MTHREAD] Got None")
return
else:
logger.debug("[MTHREAD] Received message: {}".format(msg))
task_fut = self.tasks[msg['task_id']]
if 'result' in msg:
result, _ = deserialize(msg['result'])
task_fut.set_result(result)
elif 'exception' in msg:
exception, _ = deserialize(msg['exception'])
task_fut.set_exception(exception)
if not self.is_alive:
break
def _queue_management_worker(self):
"""Listen to the queue for task status messages and handle them.
Depending on the message, tasks will be updated with results, exceptions,
or updates. It expects the following messages:
.. code:: python
{
"task_id" : <task_id>
"result" : serialized result object, if task succeeded
... more tags could be added later
}
{
"task_id" : <task_id>
"exception" : serialized exception object, on failure
}
We do not support these yet, but they could be added easily.
.. code:: python
{
"task_id" : <task_id>
"cpu_stat" : <>
"mem_stat" : <>
"io_stat" : <>
"started" : tstamp
}
The `None` message is a die request.
"""
logger.debug("[MTHREAD] queue management worker starting")
while not self.bad_state_is_set:
try:
msgs = self.incoming_q.get(timeout=1)
except queue.Empty:
logger.debug("[MTHREAD] queue empty")
# Timed out.
pass
except IOError as e:
logger.exception(
"[MTHREAD] Caught broken queue with exception code {}: {}".
format(e.errno, e))
return
except Exception as e:
logger.exception(
"[MTHREAD] Caught unknown exception: {}".format(e))
return
else:
if msgs is None:
logger.debug("[MTHREAD] Got None, exiting")
return
else:
for serialized_msg in msgs:
try:
msg = pickle.loads(serialized_msg)
tid = msg['task_id']
except pickle.UnpicklingError:
raise BadMessage(
"Message received could not be unpickled")
except Exception:
raise BadMessage(
"Message received does not contain 'task_id' field"
)
if tid == -1 and 'exception' in msg:
logger.warning(
"Executor shutting down due to exception from interchange"
)
exception = deserialize(msg['exception'])
self.set_bad_state_and_fail_all(exception)
break
task_fut = self.tasks.pop(tid)
if 'result' in msg:
result = deserialize(msg['result'])
task_fut.set_result(result)
elif 'exception' in msg:
try:
s = deserialize(msg['exception'])
# s should be a RemoteExceptionWrapper... so we can reraise it
if isinstance(s, RemoteExceptionWrapper):
try:
s.reraise()
except Exception as e:
task_fut.set_exception(e)
elif isinstance(s, Exception):
task_fut.set_exception(s)
else:
raise ValueError(
"Unknown exception-like type received: {}"
.format(type(s)))
except Exception as e:
# TODO could be a proper wrapped exception?
task_fut.set_exception(
DeserializationError(
"Received exception, but handling also threw an exception: {}"
.format(e)))
else:
raise BadMessage(
"Message received is neither result or exception"
)
if not self.is_alive:
break
logger.info("[MTHREAD] queue management worker finished")