I have a working Dask client code like that :
client = Client(address=self.cluster)
futures = []
for job in jobs:
future = client.submit(...)
futures.append(future)
for future, result in as_completed(futures, with_results=True, raise_errors=True):
key = future.key
state = (State.FINISHED if result is True else State.FAILED)
...
The Dask as_completed function is relevant, because it iterate on job that have finished with the good order.
The problem with that code, is it may block indefinitely on the as_completed call, in case of the workers are not available for instance.
Is there a way to rewrite it with asyncio ? Indeed, with asyncio, I may use the wait function with a timeout, in order to unblock blocking call, in case of errors.
Thank you
You can use asyncio.as_completed https://docs.python.org/3/library/asyncio-task.html
Related
I am trying retrieve stock prices and process the prices them as they come. I am a beginner with concurrency but I thought this set up seems suited to an asyncio producers-consumers model in which each producers retrieve a stock price, and pass it to the consumers vial a queue. Now the consumers have do the stock price processing in parallel (multiprocessing) since the work is CPU intensive. Therefore I would have multiple consumers already working while not all the producers are finished retrieving data. In addition, I would like to implement a step in which, if the consumer finds that the stock price it's working on is invalid , we spawn a new consumer job for that stock.
So far, i have the following toy code that sort of gets me there, but has issues with my process_data function (the consumer).
from concurrent.futures import ProcessPoolExecutor
import asyncio
import random
import time
random.seed(444)
#producers
async def retrieve_data(ticker, q):
'''
Pretend we're using aiohttp to retrieve stock prices from a URL
Place a tuple of stock ticker and price into asyn queue as it becomes available
'''
start = time.perf_counter() # start timer
await asyncio.sleep(random.randint(4, 8)) # pretend we're calling some URL
price = random.randint(1, 100) # pretend this is the price we retrieved
print(f'{ticker} : {price} retrieved in {time.perf_counter() - start:0.1f} seconds')
await q.put((ticker, price)) # place the price into the asyncio queue
#consumers
async def process_data(q):
while True:
data = await q.get()
print(f"processing: {data}")
with ProcessPoolExecutor() as executor:
loop = asyncio.get_running_loop()
result = await loop.run_in_executor(executor, data_processor, data)
#if output of data_processing failed, send ticker back to queue to retrieve data again
if not result[2]:
print(f'{result[0]} data invalid. Retrieving again...')
await retrieve_data(result[0], q) # add a new task
q.task_done() # end this task
else:
q.task_done() # so that q.join() knows when the task is done
async def main(tickers):
q = asyncio.Queue()
producers = [asyncio.create_task(retrieve_data(ticker, q)) for ticker in tickers]
consumers = [asyncio.create_task(process_data(q))]
await asyncio.gather(*producers)
await q.join() # Implicitly awaits consumers, too. blocks until all items in the queue have been received and processed
for c in consumers:
c.cancel() #cancel the consumer tasks, which would otherwise hang up and wait endlessly for additional queue items to appear
'''
RUN IN JUPYTER NOTEBOOK
'''
start = time.perf_counter()
tickers = ['AAPL', 'AMZN', 'TSLA', 'C', 'F']
await main(tickers)
print(f'total elapsed time: {time.perf_counter() - start:0.2f}')
'''
RUN IN TERMINAL
'''
# if __name__ == "__main__":
# start = time.perf_counter()
# tickers = ['AAPL', 'AMZN', 'TSLA', 'C', 'F']
# asyncio.run(main(tickers))
# print(f'total elapsed time: {time.perf_counter() - start:0.2f}')
The data_processor() function below, called by process_data() above needs to be in a different cell in Jupyter notebook, or a separate module (from what I understand, to avoid a PicklingError)
from multiprocessing import current_process
def data_processor(data):
ticker = data[0]
price = data[1]
print(f'Started {ticker} - {current_process().name}')
start = time.perf_counter() # start time counter
time.sleep(random.randint(4, 5)) # mimic some random processing time
# pretend we're processing the price. Let the processing outcome be invalid if the price is an odd number
if price % 2==0:
is_valid = True
else:
is_valid = False
print(f"{ticker}'s price {price} validity: --{is_valid}--"
f' Elapsed time: {time.perf_counter() - start:0.2f} seconds')
return (ticker, price, is_valid)
THE ISSUES
Instead of using python's multiprocessing module, i used concurrent.futures' ProcessPoolExecutor, which I read is compatible with asyncio (What kind of problems (if any) would there be combining asyncio with multiprocessing?). But it seems that I have to choose between retrieving the output (result) of the function called by the executor and being able to run several subprocesses in parallel. With the construct below, the subprocesses run sequentially, not in parallel.
with ProcessPoolExecutor() as executor:
loop = asyncio.get_running_loop()
result = await loop.run_in_executor(executor, data_processor, data)
Removing result = await in front of loop.run_in_executor(executor, data_processor, data) allows to run several consumers in parallel, but then I can't collect their results from the parent process. I need the await for that. And then of course the remaining of the code block will fail.
How can I have these subprocesses run in parallel and provide the output? Perhaps it needs a different construct or something else than the producers-consumers model
the part of the code that requests invalid stock prices to be retrieved again works (provided I can get the result from above), but it is ran in the subprocess that calls it and blocks new consumers from being created until the request is fulfilled. Is there a way to address this?
#if output of data_processing failed, send ticker back to queue to retrieve data again
if not result[2]:
print(f'{result[0]} data invalid. Retrieving again...')
await retrieve_data(result[0], q) # add a new task
q.task_done() # end this task
else:
q.task_done() # so that q.join() knows when the task is done
But it seems that I have to choose between retrieving the output (result) of the function called by the executor and being able to run several subprocesses in parallel.
Luckily this is not the case, you can also use asyncio.gather() to wait for multiple items at once. But you obtain data items one by one from the queue, so you don't have a batch of items to process. The simplest solution is to just start multiple consumers. Replace
# the single-element list looks suspicious anyway
consumers = [asyncio.create_task(process_data(q))]
with:
# now we have an actual list
consumers = [asyncio.create_task(process_data(q)) for _ in range(16)]
Each consumer will wait for an individual task to finish, but that's ok because you'll have a whole pool of them working in parallel, which is exactly what you wanted.
Also, you might want to make executor a global variable and not use with, so that the process pool is shared by all consumers and lasts as long as the program. That way consumers will reuse the worker processes already spawned instead of having to spawn a new process for each job received from the queue. (That's the whole point of having a process "pool".) In that case you probably want to add executor.shutdown() at the point in the program where you don't need the executor anymore.
Reading the asyncio documentation, I realize that I don't understand a very basic and fundamental aspect: the difference between awaiting a coroutine directly, and awaiting the same coroutine when it's wrapped inside a task.
In the documentation examples the two calls to the say_after coroutine are running sequentially when awaited without create_task, and concurrently when wrapped in create_task. So I understand that this is basically the difference, and that it is quite an important one.
However what confuses me is that in the example code I read everywhere (for instance showing how to use aiohttp), there are many places where a (user-defined) coroutine is awaited (usually in the middle of some other user-defined coroutine) without being wrapped in a task, and I'm wondering why that is the case. What are the criteria to determine when a coroutine should be wrapped in a task or not?
What are the criteria to determine when a coroutine should be wrapped in a task or not?
You should use a task when you want your coroutine to effectively run in the background. The code you've seen just awaits the coroutines directly because it needs them running in sequence. For example, consider an HTTP client sending a request and waiting for a response:
# these two don't make too much sense in parallel
await session.send_request(req)
resp = await session.read_response()
There are situations when you want operations to run in parallel. In that case asyncio.create_task is the appropriate tool, because it turns over the responsibility to execute the coroutine to the event loop. This allows you to start several coroutines and sit idly while they execute, typically waiting for some or all of them to finish:
dl1 = asyncio.create_task(session.get(url1))
dl2 = asyncio.create_task(session.get(url2))
# run them in parallel and wait for both to finish
resp1 = await dl1
resp2 = await dl2
# or, shorter:
resp1, resp2 = asyncio.gather(session.get(url1), session.get(url2))
As shown above, a task can be awaited as well. Just like awaiting a coroutine, that will block the current coroutine until the coroutine driven by the task has completed. In analogy to threads, awaiting a task is roughly equivalent to join()-ing a thread (except you get back the return value). Another example:
queue = asyncio.Queue()
# read output from process in an infinite loop and
# put it in a queue
async def process_output(cmd, queue, identifier):
proc = await asyncio.create_subprocess_shell(cmd)
while True:
line = await proc.readline()
await queue.put((identifier, line))
# create multiple workers that run in parallel and pour
# data from multiple sources into the same queue
asyncio.create_task(process_output("top -b", queue, "top")
asyncio.create_task(process_output("vmstat 1", queue, "vmstat")
while True:
identifier, output = await queue.get()
if identifier == 'top':
# ...
In summary, if you need the result of a coroutine in order to proceed, you should just await it without creating a task, i.e.:
# this is ok
resp = await session.read_response()
# unnecessary - it has the same effect, but it's
# less efficient
resp = await asyncio.create_task(session.read_reponse())
To continue with the threading analogy, creating a task just to await it immediately is like running t = Thread(target=foo); t.start(); t.join() instead of just foo() - inefficient and redundant.
Consider a blocking function: this_thread::sleep_for(milliseconds(3000));
I'm trying to get the following behavior:
Trigger Blocking Function
|---------------------------------------------X
I want to trigger the blocking function and if it takes too long (more than two seconds), it should timeout.
I've done the following:
my_connection = observable<>::create<int>([](subscriber<int> s) {
auto s2 = observable<>::just(1, observe_on_new_thread()) |
subscribe<int>([&](auto x) {
this_thread::sleep_for(milliseconds(3000));
s.on_next(1);
});
}) |
timeout(seconds(2), observe_on_new_thread());
I can't get this to work. For starters, I think s can't on_next from a different thread.
So my question is, what is the correct reactive way of doing this? How can I wrap a blocking function in rxcpp and add a timeout to it?
Subsequently, I want to get an RX stream that behaves like this:
Trigger Cleanup
|------------------------X
(Delay) Trigger Cleanup
|-----------------X
Great question! The above is pretty close.
Here is an example of how to adapt blocking operations to rxcpp. It does libcurl polling to make http requests.
The following should do what you intended.
auto sharedThreads = observe_on_event_loop();
auto my_connection = observable<>::create<int>([](subscriber<int> s) {
this_thread::sleep_for(milliseconds(3000));
s.on_next(1);
s.on_completed();
}) |
subscribe_on(observe_on_new_thread()) |
//start_with(0) | // workaround bug in timeout
timeout(seconds(2), sharedThreads);
//skip(1); // workaround bug in timeout
my_connection.as_blocking().subscribe(
[](int){},
[](exception_ptr ep){cout << "timed out" << endl;}
);
subscribe_on will run the create on a dedicated thread, and thus create is allowed to block that thread.
timeout will run the timer on a different thread, that can be shared with others, and transfer all the on_next/on_error/on_completed calls to that same thread.
as_blocking will make sure that subscribe does not return until it has completed. This is only used to prevent main() from exiting - most often in test or example programs.
EDIT: added workaround for bug in timeout. At the moment, it does not schedule the first timeout until the first value arrives.
EDIT-2: timeout bug has been fixed, the workaround is not needed anymore.
I am downloading jsons from an api and am using the asyncio module. The crux of my question is, with the following event loop as implemented as this:
loop = asyncio.get_event_loop()
main_task = asyncio.ensure_future( klass.download_all() )
loop.run_until_complete( main_task )
and download_all() implemented like this instance method of a class, which already has downloader objects created and available to it, and thus calls each respective download method:
async def download_all(self):
""" Builds the coroutines, uses asyncio.wait, then sifts for those still pending, loops """
ret = []
async with aiohttp.ClientSession() as session:
pending = []
for downloader in self._downloaders:
pending.append( asyncio.ensure_future( downloader.download(session) ) )
while pending:
dne, pnding= await asyncio.wait(pending)
ret.extend( [d.result() for d in dne] )
# Get all the tasks, cannot use "pnding"
tasks = asyncio.Task.all_tasks()
pending = [tks for tks in tasks if not tks.done()]
# Exclude the one that we know hasn't ended yet (UGLY)
pending = [t for t in pending if not t._coro.__name__ == self.download_all.__name__]
return ret
Why is it, that in the downloaders' download methods, when instead of the await syntax, I choose to do asyncio.ensure_future instead, it runs way faster, that is more seemingly "asynchronously" as I can see from the logs.
This works because of the way I have set up detecting all the tasks that are still pending, and not letting the download_all method complete, and keep calling asyncio.wait.
I thought that the await keyword allowed the event loop mechanism to do its thing and share resources efficiently? How come doing it this way is faster? Is there something wrong with it? For example:
async def download(self, session):
async with session.request(self.method, self.url, params=self.params) as response:
response_json = await response.json()
# Not using await here, as I am "supposed" to
asyncio.ensure_future( self.write(response_json, self.path) )
return response_json
async def write(self, res_json, path):
# using aiofiles to write, but it doesn't (seem to?) support direct json
# so converting to raw text first
txt_contents = json.dumps(res_json, **self.json_dumps_kwargs);
async with aiofiles.open(path, 'w') as f:
await f.write(txt_contents)
With full code implemented and a real API, I was able to download 44 resources in 34 seconds, but when using await it took more than three minutes (I actually gave up as it was taking so long).
When you do await in each iteration of for loop it will await to download every iteration.
When you do ensure_future on the other hand it doesn't it creates task to download all the files and then awaits all of them in second loop.
I've the following problem that is begging a zmq solution. I have a time-series data:
A,B,C,D,E,...
I need to perform an operation, Func, on each point.
It makes good sense to parallelize the task using multiple workers via zmq. However, what is tripping me up is how do I synchronize the result, i.e., the results should be time-ordered exactly the way the input data came in. So the end result should look like:
Func(A), Func(B), Func(C), Func(D),...
I should also point out that time to complete,say, Func(A) will be slightly different than Func(B). This may require me to block for a while.
Any suggestions would be greatly appreciated.
You will always need to block for a while in order to synchronize things. You can actually send requests to a pool of workers, and when a response is received - to buffer it if it is not a subsequent one. One simple workflow could be described in a pseudo-language as follows:
socket receiver; # zmq.PULL
socket workers; # zmq.DEALER, the worker thread socket is started as zmq.DEALER too.
poller = poller(receiver, workers);
next_id_req = incr()
out_queue = queue;
out_queue.last_id = next_id_req
buffer = sorted_queue;
sock = poller.poll()
if sock is receiver:
packet_N = receiver.recv()
# send N for processing
worker.send(packet_N, ++next_id_req)
else if sock is workers:
# get a processed response Func(N)
func_N_response, id = workers.recv()
if out_queue.last_id != id-1:
# not subsequent id, buffer it
buffer.push(id, func_N_rseponse)
else:
# in order, push to out queue
out_queue.push(id, func_N_response)
# also consume all buffered subsequent items
while (out_queue.last_id == buffer.min_id() - 1):
id, buffered_N_resp = buffer.pop()
out_queue.push(id, buffered_N_resp)
But here comes the problem what happens if a packet is lost in the processing thread(the workers pool).. You can either skip it after a certain timeout(flush the buffer into the out queue), amd continue filling the out queue, and reorder when the packet comes later, if ever comes.