gh-75880: Deadlocks in concurrent.futures.ProcessPoolExecutor with unpickling error#4256
Closed
tomMoral wants to merge 3 commits into
Closed
gh-75880: Deadlocks in concurrent.futures.ProcessPoolExecutor with unpickling error#4256tomMoral wants to merge 3 commits into
concurrent.futures.ProcessPoolExecutor with unpickling error#4256tomMoral wants to merge 3 commits into
Conversation
f72d0fd to
51b411d
Compare
Contributor
|
The windows failures are unrelated to this PR ( |
51b411d to
4cf2a47
Compare
4cf2a47 to
a70485b
Compare
tomMoral
commented
Dec 1, 2017
a70485b to
b7422fe
Compare
ogrisel
suggested changes
Jan 12, 2018
ogrisel
left a comment
Contributor
There was a problem hiding this comment.
I don't what the best solution is but this PR should not be merged while the following regression is not addressed.
| work_id, exception=_ExceptionWithTraceback(e, e.__traceback__) | ||
| )) | ||
| result_queue._put_bytes(work_id.to_bytes(WORK_ID_SIZE, WORK_ID_ENC) + | ||
| serialize_res) |
Contributor
There was a problem hiding this comment.
This bytes concatenation triggers a potentially very large memory copy which is a regression compared to master.
There are two possible solutions to this problem:
- Add a
send_byte_chunksAPI down to the Connection that would make it possible to send a topple of bytes without materializing the full buffer. We don't know if it is possible on windows. - Implement a new Queue sub class to send and receive 2 messages at a time. This would introduce a small call overhead, potentially critical for small messages.
concurrent.futures.ProcessPoolExecutor with unpickling errorconcurrent.futures.ProcessPoolExecutor with unpickling error
Member
|
https://bugs.python.org/issue31699 is closed. What is the status of this PR? |
|
This PR is stale because it has been open for 30 days with no activity. |
concurrent.futures.ProcessPoolExecutor with unpickling errorconcurrent.futures.ProcessPoolExecutor with unpickling error
|
This PR is stale because it has been open for 30 days with no activity. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
8 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
When using
concurrent.futures.ProcessPoolExecutorwith objects that are not picklable or unpicklable, several situations results in a deadlock, with the interpreter frozen.This is the case for different scenarios, for instance, https://gist.github.com/tomMoral/cc27a938d669edcf0286c57516942369. This PR is a follow up of #3895 and specifically fixes the unpickling behavior. With this PR, the unpickling failures do not result in broken executors. This is done by ensuring that the
_callItemand the_ResultItemare sent in queues along the work_id with a specific communication protocol, which sends thework_idas bytes, followed by the object serialized bypickle.To this end, we introduce private API in
multiprocessing.Queueto allow modified serialization protocols.Overall, the goal is to make
concurrent.futures.ProcessPoolExecutormore robust to faulty user code.This work was done as part of the tommoral/loky#48 with the intent to re-use the executor in multiple independent parts of the program, in collaboration with @ogrisel. See #1013 for more the details.
https://bugs.python.org/issue31699
concurrent.futures.ProcessPoolExecutorwith pickling error #75880