Switch drain_all call to be called from new thread instead of event loop.#16739
Conversation
CodSpeed Performance ReportMerging #16739 will not alter performanceComparing Summary
|
…th-dependencies-that
There was a problem hiding this comment.
I still seem to remember some issue about spawning a new thread at interpreter shutdown a la python/cpython#116677 but maybe this has since been fixed and released in a recent version of python 3.12 cc @jlowin (I remember we both ran into this)
anyways if we're passing tests and not seeing any issues, I think this makes sense
src/prefect/logging/handlers.py
Outdated
| @@ -118,7 +118,7 @@ def flush(cls) -> None: | |||
|
|
|||
| # Not ideal, but this method is called by the stdlib and cannot return a | |||
| # coroutine so we just schedule the drain in the global loop thread and continue | |||
There was a problem hiding this comment.
comment should probably be updated to match new behavior
| # coroutine so we just schedule the drain in the global loop thread and continue | |
| # coroutine so we just schedule the drain in a new thread and continue |
…cies-that' of https://github.com/PrefectHQ/prefect into jean/oss-5963-deadlock-when-deloying-flow-with-dependencies-that
…th-dependencies-that
| logger = logging.getLogger(__name__) | ||
| logger.setLevel("WARNING") | ||
|
|
||
| dictConfig({"version": 1, "incremental": False}) |
There was a problem hiding this comment.
the bug ticket says the bug is triggered with incremental=True?
There was a problem hiding this comment.
That was my mistake, it is incremental=False as that is what makes configure call _clearExistingHandlers
jeanluciano
deleted the
jean/oss-5963-deadlock-when-deloying-flow-with-dependencies-that
branch
3 participants
Moves the
drain_allcall to a new thread instead of the event loop.closes: #16115
Checklist
<link to issue>"mint.json.