Avoid setting event loop policy if within Jupyter notebook server #2343
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
This is a hack. If Dask is imported within the Jupyter notebook server it can cause the server to hang. This occurs because Dask sets an asyncio event loop policy useful when starting multiple event loops in multiple threads. Unfortunately setting this policy after Jupyter has already created an event loop causes Jupyter to have two different event loops, which understandably causes difficulties. In the future we should investigate avoiding setting global asyncio policies, possibly through managing event loops ourselves without using `asyncio.get_event_loop` but until then this special-cased hack should relieve some pressure. Note that this means that some advanced functionality, like `get_client`, `as_completed` and so forth won't work. Fortunately these are unlikely to be relevant within the Jupyter server process, which is more likely to run scheduler rather than client code. See jupyter/notebook#4183
minrk
approved these changes
Nov 8, 2018
mrocklin
pushed a commit
that referenced
this pull request
Dec 31, 2019
…ng event loop (#3336) Setting asyncio event loop policy at these two places could cause problems. 1. When policy is set in Jupyter notebook server extension. This causes the notebook server to hang. This is fixed in #2343. 2. When policy is set in iPython startup config (`~/.ipython/profile_default/startup/whatever.py`) or by setting `get_config().InteractiveShellApp.exec_lines` in `~/.ipython/profile_default/ipython_config.py`. This causes the kernel to hang. This can be reproduced by running either `jupyter console` or `jupyter notebook`. If running In Jupyter notebook, it will struck at "Kernel starting, please wait...". Note that manually setting the policy in notebook cell, after the kernel has started, is fine. In both cases, running `asyncio.get_running_loop()` just before setting the policy will raise `RuntimeError`, meaning there is no running event loop yet. See #3202.
bnaul
pushed a commit
to replicahq/distributed
that referenced
this pull request
Feb 10, 2020
…ng event loop (dask#3336) Setting asyncio event loop policy at these two places could cause problems. 1. When policy is set in Jupyter notebook server extension. This causes the notebook server to hang. This is fixed in dask#2343. 2. When policy is set in iPython startup config (`~/.ipython/profile_default/startup/whatever.py`) or by setting `get_config().InteractiveShellApp.exec_lines` in `~/.ipython/profile_default/ipython_config.py`. This causes the kernel to hang. This can be reproduced by running either `jupyter console` or `jupyter notebook`. If running In Jupyter notebook, it will struck at "Kernel starting, please wait...". Note that manually setting the policy in notebook cell, after the kernel has started, is fine. In both cases, running `asyncio.get_running_loop()` just before setting the policy will raise `RuntimeError`, meaning there is no running event loop yet. See dask#3202.
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.
This is a hack.
If Dask is imported within the Jupyter notebook server it can cause the
server to hang. This occurs because Dask sets an asyncio event loop
policy useful when starting multiple event loops in multiple threads.
Unfortunately setting this policy after Jupyter has already created an
event loop causes Jupyter to have two different event loops, which
understandably causes difficulties.
In the future we should investigate avoiding setting global asyncio
policies, possibly through managing event loops ourselves without using
asyncio.get_event_loopbut until then this special-cased hack shouldrelieve some pressure.
Note that this means that some advanced functionality, like
get_client,as_completedand so forth won't work. Fortunately theseare unlikely to be relevant within the Jupyter server process, which is
more likely to run scheduler rather than client code.
See jupyter/notebook#4183