Sign in

My Ruby Job Runner Segfaults After The New OS Image And The Logs Say NOTHING

debugginglinuxsegfault13 replies · 495 views · started
Reply
  1. Joon Quintero
    @joon_ping

    guys pls help reproducible crash, at least... the new OS image kills teh Ruby job runner instantly:

    admin@box:~/build$ ./report-tool --input data.csv

    admin@box:~/build$ ./report-tool --input data.csv
loading 48213 rows...
parsing columns... ok
building index...
Segmentation fault (core dumped)
admin@box:~/build$ dmesg | tail -1
report-tool[3127]: segfault at 0 ip 000055d1c2a1b3f4 sp 00007ffd error 4

    i found a core file somewhere once but no idea what to do with it. Do i really need to learn gdb tonihgt?

    I can reproduce it in staging with the exact same input, which feels like a gift. I just need to know which tools to point at it. Core dumps, debug flags, tracing, whatever. Teach me how to fish here, please. any ideas??

    28
  2. Yusuf Fischer
    @cron_garage483

    wait did heavy concurrency also change any native libraries (livbips, ImageMagick, libxml2, OpenSSL)?

    1 0 helpful
  6. Lea Dalton
    @stacksmith-vo
    in reply to @cron_garage483

    oh ive been there lol if it's concurrency-related, look for a non-thread-safe library being shared across threads. Running teh worker with one thread temporarily is a quick way to confirm.

    2 1 helpful
  7. Elena E.
    @eeriksen-1wq

    high-level code segfaults through native dependencies: image libraries, database drivers, XML parsers, crypto bindings. after heavy concurrency, check whether a native library changed version underneath a compiled extension. rebuilding extensions against the current system libs (pip install --no-binary / npm rebuild) often fixes it.

    2 0 helpful
  8. Esme Schmidt
    @eschmidt-2ss

    lol same thing happened to me get a core dump first. on systemd distros theyre captured by systemd-coredump:

    coredumpctl list
coredumpctl info <PID>
coredumpctl gdb <PID>

    in gdb, bt prints the backtrace... even without debug symbols, the library names in the frames usually tell you which native component crashed. fingers crossed 🤞

    1 0 helpful
  9. Elena E.
    @eeriksen-1wq
    in reply to @eschmidt-2ss

    been there man if you're in a container, core dumps may be disabled... check ulimit -c (set it to unlimited) and cat /proc/sys/kernel/core_pattern on the host, since the pattern is host-wide. hope that helps :)

    17 0 helpful
  11. Soren Fontaine
    @soren.fontaine
    in reply to @cziegler

    hmmm for Python, add import faulthandler; faulthandler.enable() (or run with python -X faulthandler). On a segfault it prints the Python-level traceback of every thread, which shows which call into the C extension crashed. worked for me anyway 🤷

    2 0 helpful
  12. Sami B.
    @solidfiddler

    wait which version are you on? ??

    0 6 helpful
  13. Lea Dalton
    @stacksmith-vo

    hmm idk before gdb, Id just pin the previous base image and confirm the crash disappears. bisecting is faster than reading assembly.

    2 5 helpful
  14. Joon Quintero
    @joon_ping
    in reply to @soren.fontaine

    @soren.fontaine omggg faulthandler showed the exact Python line... it was a thread-unsafe native call... wrapped it in a lock and the segfaults stopped. thank u thank u thank u

    2 14 helpful

Reply to this topic

Sign in or create an account to reply.