MOTOR-1087 motor_asyncio freezes when using multiprocessing #196
Conversation
motor/frameworks/asyncio/__init__.py
Outdated
| if hasattr(os, "register_at_fork"): | ||
| # This will run in the same thread as the fork was called. | ||
| # If we fork in a critical region on the same thread, it should break. | ||
| # This is fine since we would never call fork directly from a critical region. |
There was a problem hiding this comment.
What is this comment trying to say?
There was a problem hiding this comment.
I copied this over from Ben's PR, which had this comment for all of the register_at_fork calls. I can remove if you think it is not valuable?
There was a problem hiding this comment.
Yeah I don't think its valuable. The problem is that it's only describing what after_in_child does. If there was a comment here it should explain why we call it, eg "ThreadPoolExecutor is not fork safe, reinitialize it post fork."
There was a problem hiding this comment.
Should be fixed now.
| child_exec = parent_conn.recv() | ||
| self.assertIsNot(child_exec, init_exec) | ||
| passed, msg = parent_conn.recv() | ||
| self.assertTrue(passed, msg) |
There was a problem hiding this comment.
Instead of testing that the id of asyncio._EXECUTOR changes, we should test that a new motor client can actually be used in the child process.
motor/frameworks/asyncio/__init__.py
Outdated
|
|
||
|
|
||
| if hasattr(os, "register_at_fork"): | ||
| # We need this to make sure that creating new clients in threads doesn't deadlock. |
(cherry picked from commit 935c543)
No description provided.