[issue35988] Python interpreter segfault

Hinko Kocevar added the comment: Thank you for the valuable information about the OOM beast.. -- ___ Python tracker ___ ___

[issue35988] Python interpreter segfault

Josh Rosenberg added the comment: "your application is using more memory than what is available in the system." Well, it alone may not be using more memory, but the cumulative usage on the system is "too high" by whatever metric the OOM killer is using (IIRC the default rule is that actual

[issue35988] Python interpreter segfault

Hinko Kocevar added the comment: Using python3.6 now. I've also used the 'handle SIGPIPE nostop noprint' to skip SIGPIPE. Now the interpreter is killed instead: INF 2019-02-13 15:42:19,131 web:2162 log_request - 200 GET /NPM.html (10.0.6.209) 1.10ms [Thread 0x7fffe76eb700 (LWP 8411)

[issue35988] Python interpreter segfault

STINNER Victor added the comment: > [26194250.314042] Out of memory: Kill process 8399 (python3.6) score 574 or > sacrifice child That's not a bug in Python: your application is using more memory than what is available in the system. -- ___

[issue35988] Python interpreter segfault

Hinko Kocevar added the comment: > Are you able to reproduce the crash on Python 3.6 or 3.7? I'm running it under python3.6 as we speak. -- ___ Python tracker ___

[issue35988] Python interpreter segfault

STINNER Victor added the comment: > Every now and then the python3.5 crashes, seg faults. Sorry but Python 3.5 no longer accepts bugfixes and has known bugs. http://devguide.python.org/#status-of-python-branches It's time to upgrade your Python! Are you able to reproduce the crash on Python

[issue35988] Python interpreter segfault

STINNER Victor added the comment: You can enable faulthandler to get a traceback of your Python theads when a crash occurs: https://pythondev.readthedocs.io/debug_tools.html#get-a-traceback-on-a-crash -- ___ Python tracker

[issue35988] Python interpreter segfault

STINNER Victor added the comment: > Program received signal SIGPIPE, Broken pipe. SIGPIPE is not a bug, it's a deliberate signal to notice something to your application. You have to ignore SIGPIPE in gdb: (gdb) handle SIGPIPE nostop noprint SignalStop Print Pass to program

[issue35988] Python interpreter segfault

New submission from Hinko Kocevar : I'm running a tornado server with websockets client. Every now and then the python3.5 crashes, seg faults. I added code tracking (https://stackoverflow.com/questions/2663841/python-tracing-a-segmentation-fault) and this is what I see: line,