I have a notebook using tensorflow with pandas which consistently restarts the kernel in a specific cell. The same notebook runs fine on my laptop.
How can I diagnose what is killing the kernel in SWAN?
Hi David,
You can try to run the same thing from the SWAN terminal to check if you get a meaningful error.
Otherwise if you let me know which node you logged in to I can inspect the logs of your session.
Cheers,
Enric
1 Like
It can also be that you get out of memory somehow. You can try increasing notebook memory setting before Connecting.
Dear SWAN team,
In order to continue on the subject, I encountered some instabilities with my kernel on swan006 (connection lost or failed), around 15h00 today if this can help…the connection to swan itself (hub/browser) was somehow slow (but I do not know if this is on my side only or not…). Moreover I only had 2 running notebooks in parallel, neither of them asking huge resources.
Bests,
Brian
Dear Brian,
Sorry for the late reply, did the connection failed error prevent you from using the notebook afterwards? Sometimes you see that error but the kernel is still ok.
Dear Enric,
I had to refresh the notebook and it worked after…But some times it says the kernel is running (top right) but nothing cell is working anymore (no *) and any ask for running a cell is immediatly “killed” without output. In that case I have to exit the notebook, kill it the hard way through the web interface and then reopen it.
Anyway this is not really helpful…I will try to make better reports today if I have any new issue (and remember better what I did then)
Bests,
Brian