equinor
No running event loop error observed when doing heavy update
Observed while running the poly-case on version 9.0.4 on 1000 realizations, during the update step:
(2024.04.rc0-py38) [havb@be-linrgsn001:~/projects/ert/test-data/poly_example] main$ ert gui --enable-scheduler poly.ert
Exception ignored in: <coroutine object WebSocketCommonProtocol.close_connection at 0x7f2c8c56c340>
Traceback (most recent call last):
File "/prog/res/komodo/2024.04.rc0-py38-rhel7/root/lib64/python3.8/site-packages/websockets/legacy/protocol.py", line 1337, in close_connection
await self.close_transport()
File "/prog/res/komodo/2024.04.rc0-py38-rhel7/root/lib64/python3.8/site-packages/websockets/legacy/protocol.py", line 1355, in close_transport
if await self.wait_for_connection_lost():
File "/prog/res/komodo/2024.04.rc0-py38-rhel7/root/lib64/python3.8/site-packages/websockets/legacy/protocol.py", line 1379, in wait_for_connection_lost
async with asyncio_timeout(self.close_timeout):
File "/prog/res/komodo/2024.04.rc0-py38-rhel7/root/lib64/python3.8/site-packages/websockets/legacy/async_timeout.py", line 74, in timeout
loop = asyncio.get_running_loop()
RuntimeError: no running event loop
This is probably the Ensemble Evaluator that is too busy doing the maths for the update from an iteration to the next, and not setting aside time to maintain the websocket connection, and then some bad handling of this problem.
The poly case progresses fine so this is not an Error, just a Warning. But this should not be shown to users.
Now I start to understand why this happens. Not sure if this is an inherit flakiness of checking self._connection.openwhich should be the proper way of testing it before calling connection.close() in monitor. However, when having this extra check then all hell got loose: https://github.com/equinor/ert/actions/runs/9284739204/job/25547764539#step:7:541
