Error when running on AWS Lambda

[DaaS-20xx]

Hi all, everything is working fine locally, but when deployed on AWS Lambda, the simple_file.py application doesn't work. Here the steps I executed:

• I created a new Lambda function, assigning Role with policy permission to access Transcribe service
• I changed the default handler to "simple_file.basic_transcribe" (simple_file.py is the application in the example folder and basic_transcribe is the main function inside).
• Locally I created a "deploy-transcribe" folder, and within it I placed the simple_file.py; I've also modified the basic_transcribe function to include the parameters event and context (to make the simple_file.py the lambda handler). I also placed the sample file.wav in S3 bucket and updated the address in the aiofile command in the simple_file.py.
• in the same local folder, I installed the dependencies amazon-transcribe and aiofile via "pip -t ."; I also installed manually the awscrt by downloading from here: https://pypi.org/project/awscrt/#files and unzipping in the same local folder (as explained in the issue #18 here.
• finally I zipped the content of the entire folder, uploaded in lambda, successfully deployed.

But when testing, I get an error: "[Errno 38] Function not implemented"; here below (*) the log from CloudWatch.

Any Idea of what is the reason? Am I missing anything or should the simple_file.py script be adapted/modified in some way?

Thanks a lot!! Bye!

(*) [ERROR] OSError: [Errno 38] Function not implemented Traceback (most recent call last):   File "/var/lang/lib/python3.9/importlib/init.py", line 127, in import_module     return _bootstrap._gcd_import(name[level:], package, level)   File "", line 1030, in _gcd_import   File "", line 1007, in _find_and_load   File "", line 986, in _find_and_load_unlocked   File "", line 680, in _load_unlocked   File "", line 850, in exec_module   File "", line 228, in _call_with_frames_removed   File "/var/task/simple_file.py", line 53, in     loop.run_until_complete(basic_transcribe())   File "/var/lang/lib/python3.9/asyncio/base_events.py", line 642, in run_until_complete     return future.result()   File "/var/task/simple_file.py", line 50, in basic_transcribe     await asyncio.gather(write_chunks(), handler.handle_events())   File "/var/task/simple_file.py", line 42, in write_chunks     async with aiofile.AIOFile('s3://transcr12092021/test/test.wav', 'rb') as afp:   File "/var/task/aiofile/aio.py", line 124, in init     self.context = context or get_default_context()   File "/var/task/aiofile/aio.py", line 307, in get_default_context     return create_context()   File "/var/task/aiofile/aio.py", line 294, in create_context     context = caio.AsyncioContext(max_requests, loop=loop)   File "/var/task/caio/asyncio_base.py", line 22, in init     self.context = self._create_context(max_requests, kwargs)   File "/var/task/caio/asyncio_base.py", line 25, in _create_context     return self.CONTEXT_CLASS(max_requests=max_requests, kwargs)   File "/var/task/caio/python_aio.py", line 34, in init__     self.pool = ThreadPool(pool_size)   File "/var/lang/lib/python3.9/multiprocessing/pool.py", line 927, in init     Pool.init(self, processes, initializer, initargs)   File "/var/lang/lib/python3.9/multiprocessing/pool.py", line 196, in init     self._change_notifier = self._ctx.SimpleQueue()   File "/var/lang/lib/python3.9/multiprocessing/context.py", line 113, in SimpleQueue     return SimpleQueue(ctx=self.get_context())   File "/var/lang/lib/python3.9/multiprocessing/queues.py", line 342, in init     self._rlock = ctx.Lock()   File "/var/lang/lib/python3.9/multiprocessing/context.py", line 68, in Lock     return Lock(ctx=self.get_context())   File "/var/lang/lib/python3.9/multiprocessing/synchronize.py", line 162, in init     SemLock.init(self, SEMAPHORE, 1, 1, ctx=ctx)   File "/var/lang/lib/python3.9/multiprocessing/synchronize.py", line 57, in init     sl = self._semlock = _multiprocessing.SemLock(

[DaaS-20xx]

Figured out that, according to the details provided here: https://aws.amazon.com/blogs/compute/parallel-processing-in-python-with-aws-lambda/, it's not possible to implement parallelism on Lambda by using the Multiprocessing.Pool and Multiprocessing.Queue (from the multiprocessing module only Pipe can be used). And from the stack trace, the multiprocessing.pool is invoked when the aiofile.AIOFile is fired. Is there an alternative to that?

[DaaS-20xx]

Hi everyone, I just figured out how to fix the issue. As explained here: https://aws.amazon.com/blogs/compute/parallel-processing-in-python-with-aws-lambda/ AWS Lambda does not support multiprocessing.pool, which is used in the aiofile library (exactly multiprocessing.pool.ThreadPool is invoked in the caio.python_aio.py), and this cause the error I reported above. Now I found the same issues in this other discussion on AWS Lambda not working with snowflake connector: https://github.com/snowflakedb/snowflake-connector-python/issues/287#. In this case the issue has been fixed with the release of new version of the snowflake-connector-python library, specifically by simply replacing multiprocessing.pool.ThreadPool with concurrent.futures.thread.ThreadPoolExecutor. And I applied exactly the same fix manually to the python_aio.py within the caio folder, i.e.:

from concurrent.futures.thread import ThreadPoolExecutor .. self.pool = ThreadPoolExecutor(pool_size)

Then I zipped everything and redeployed in AWS Lambda function, and issue is fixed!

Anyway I assume that this fix should be addressed officially, e.g. some specific add-on for AWS Lambda.

Thanks everybody!

[DaaS-20xx]

Unfortunately, moving forward, the issue showed up when running the python script in lambda during the opening of the file, and it's on the ThreadPoolExecutor which I had put in place of ThreadPool in the caio.python_aio,py:

[ERROR] AttributeError: 'ThreadPoolExecutor' object has no attribute 'apply_async' Traceback (most recent call last): File "/var/task/lambda_function.py", line 61, in lambda_handler return loop.run_until_complete(basic_transcribe()) File "/var/lang/lib/python3.9/asyncio/base_events.py", line 642, in run_until_complete return future.result() File "/var/task/lambda_function.py", line 51, in basic_transcribe await asyncio.gather(write_chunks(), handler.handle_events()) File "/var/task/lambda_function.py", line 45, in write_chunks async for chunk in reader: File "/var/task/aiofile/utils.py", line 77, in anext chunk = await self.read_chunk() File "/var/task/aiofile/utils.py", line 64, in read_chunk chunk = await self.file.read_bytes( File "/var/task/aiofile/aio.py", line 220, in read_bytes return await self.__context.read(size, self.fileno(), offset) File "/var/task/caio/asyncio_base.py", line 47, in submit assert self.context.submit(op) == 1, "Operation was not submitted" File "/var/task/caio/python_aio.py", line 128, in submit self._execute(operation) File "/var/task/caio/python_aio.py", line 69, in _execute self.pool.apply_async(

So it seems the issue is with opening file in async/non blocking mode in AWS Lambda, that is apparently on this piece of code:

async with aiofile.AIOFile('/tmp/test.wav', 'rb') as afp:

Any suggestion or hint on how can it be addressed? or is it not feasible in AWS Lambda?

Thanks all!