Kernel dies
After executing about 10 to 12 commands, the kernel dies with the following error.
Unhandled signal received.
Killed by SIGSEGV.
Let me know what other information you need.
I'm also running into this issue quite frequently. Running inside Docker (2.3.0.3) on a MacOS Catalina (10.15.4).
Would love to take a shot on this problem, but I have no idea where to start...
I spent a few hours digging around this weekend - this is a tough one. It's happening consistently for me around every 8th or 9th command.
It's definitely not something that used to happen, which makes me suspect it has something to do with changes to the jupyter kernel spec or ZMQ. Maybe the bindings in zmq.cdecl are no longer accurate?
I tried to track down exactly where in the execution pipeline the segfault happens and it seems like it happens after everything related to the execution request completes. In order, the kernel
- receives a is_complete_request
- signals that it's busy
- responds with an is_complete_reply
- signals that it's idle
- receives an execute request
- evaluates the expression
- responds with an execution reply
- signals that it's idle
- and then segfault!
I'll keep digging around; just thought I'd post these updates.
I'm facing a similar problem on MacOS Mojave after running 5-6 cells.