BUG: Kill subprocesses on shutdown. #869
Merged
Conversation
This file contains 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
Fixes #jupyter/jupyter_client#104 This should make sure we properly cull all subprocesses at shutdown, it does change one of the private method from sync to async in order to no user time.sleep or thread so this may affect subclasses, though I doubt it. It's also not completely clear to me whether this works on windows as SIGINT I belove is not a thing. Regardless as this affects things like dask, and others that are mostly on unix, it should be an improvement. It does the following, stopping as soon as it does not find any more children to current process. - Send sigint to everything - Immediately send sigterm in look with an exponential backoff from 0.01 to 1 second roughtly multiplying the delay until next send by 3 each time. - Switch to sending sigkill with same backoff. There is no delay after sigint, as this is just a courtesy. The delays backoff are not configurable. I can imagine that on slow systems it may make sens
Carreau
force-pushed
the
kill-sp
branch
2 times, most recently
from
1b53afb
to
b3e8b85
Compare
blink1073
reviewed
Feb 17, 2022
blink1073
reviewed
Feb 17, 2022
Co-authored-by: Steven Silvester <[email protected]>
blink1073
reviewed
Feb 17, 2022
ipykernel/kernelbase.py
Outdated
| @@ -1131,9 +1136,62 @@ def _input_request(self, prompt, ident, parent, password=False): | |||
| raise EOFError | |||
| return value | |||
|
|
|||
| def _at_shutdown(self): | |||
| async def _progressively_terminate_all_children(self): | |||
| if sys.platform == "win32": | |||
There was a problem hiding this comment.
I think this method should work on both Windows and Posix: https://stackoverflow.com/a/25134985
There was a problem hiding this comment.
I'll give it a try. I know that the "loop and kill children" was switched back to killpg at some point in jupyter_client as the first one had issues.
There was a problem hiding this comment.
(that was in jupyter/jupyter_client#743, i'll try to get the same logic from there)
|
Kicking to make sure we hit all downstream tests |
|
Weird. I have no idea why the |
Thanks to you for the review and everything. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Fixes jupyter/jupyter_client#104.
This should make sure we properly cull all subprocesses at shutdown,
it does change one of the private method from sync to async in order to
no user time.sleep or thread so this may affect subclasses, though I
doubt it.
It's also not completely clear to me whether this works on windows as
SIGINT I belove is not a thing.
Regardless as this affects things like dask, and others that are mostly
on unix, it should be an improvement.
It does the following, stopping as soon as it does not find any more
children to current process.
0.01 to 1 second roughtly multiplying the delay until next send by 3
each time.
There is no delay after sigint, as this is just a courtesy.
The delays backoff are not configurable. I can imagine that on slow
systems it may make sens