Closed ClaytonAstrom closed 4 months ago
Maybe there's something I've missed, this is a part of our upgrade from 0.4 to 0.5. The notebooks run fine in 0.4.
Something I've noticed is that server side in 0.4 says there's a websocket log - Initializing websocket connection /api/kernels/...
However, even on a successful run in 0.5, it skips this and prints the immediate next log - Requesting kernel info from...
Hi, can you post a simplified notebook to reproduce the issue? Might be related to https://github.com/voila-dashboards/voila/issues/1428#issuecomment-1927025620 ?
Thanks @trungleduc - I took the simplified notebook you posted in that issue and I can see it seems to be the same issue. It's not consistent 100% of the time, that's also in line with the notebooks I'm testing as well. But the logs and error seem to match the notebooks I'm struggling with, so I'm willing to bet it's the same issue.
I'll close this one out then and keep an eye on the issue you linked.
Description
I'm noticing that some notebooks finish executing server side, while the front end hasn't received anything (stays on the spinner and continues sending the heartbeat). Eventually the kernel is culled due to inactivity and the front end believes the kernel died erroneously because of that. I've tried increasing kernel polling/culling timeouts, but it seems that if it's missed it's missed. Server side these logs look the same, except for when next the front end runs the heartbeat:
I'm not sure why it would say there's a server side error though despite the execution finishing above and the kernel going idle.
I should also mention this is very inconsistent, even for a specific notebook. Some notebooks never have this error, some sometimes, and others frequently (still not 100% of the time though). Which to me says something is behaving incorrectly with the async part.
Context
Troubleshoot Output
Command Line Output
Browser Output