## Debugging A Hung (Spinning) Python Process

From: andrew cooke <andrew@...>

Date: Mon, 31 May 2010 09:46:01 -0400

I left a Python script running at the client's, overnight, expecting
everything to be ready in the morning.  But when I checked next day the CPU
was pegged at 100% and the log indicated no progress since 1am.  Clearly the
Python code had a bug and was "spinning" in a tight loop.

The problem, then, was to find out what code was being executed.  Luckily
there's a very neat solution, that I've used before, that does exactly this.
It's described in an answer by "spiv" at
http://stackoverflow.com/questions/132058/getting-stack-trace-from-a-running-python-application

That describes how to connect gdb to the process ("gdb -p PID") and dump a
Python stack trace ("pystack").  All that you need to do is create a .gdbinit
file with the contents given here -
http://svn.python.org/projects/python/trunk/Misc/gdbinit

With a stack trace it was easy to see the error in my code - and it was a
tight loop, and it depended explicitly on the time of day, which is why I had
not seen it before. :o)

Andrew

### Update

From: andrew cooke <andrew@...>

Date: Thu, 31 May 2018 21:16:09 -0400

This is now simpler.

You need to follow the instructions at
https://devguide.python.org/gdb/ - basically put

in ~/.gdbinit

Then connect with gdb -p PID and use the commands described in the
Andrew