Skip to content

Latest commit

 

History

History
44 lines (31 loc) · 1.63 KB

core-dumps.md

File metadata and controls

44 lines (31 loc) · 1.63 KB
Raw

Core dumps

What is core dump?

When your process crashes, it produces a core dump. It's a snapshot of the state of your app (memory, call stack) on it's death bed.

If you didn't encounter a core dump before, it's probably because your operating system is configured to limit dump file sizes to 0 bytes by default.

Enabling core dumps

ulimit -c unlimited

Change the limit of core dump size in your linux OS.

This can also be configured with any Ops tools for infrastructure automation. Setting it to unlimited in production is not a great idea, especially if auto-restart is enabled.

If you're running in containers, make sure they're configured not to discard the storage location where dumps are written.

Core dump analysis

asciicast

  • install lldb in your system (from apt)
  • install llnode according to readme llnode notes
  • enable writing core dumps ulimit -c unlimited
  • run the app and see it break
    • node dumps when it runs out of memory or segfaults
    • set --abort-on-uncaught-exception to dump core on crash
    • call process.abort()
    • or just use gcore <PID> to generate core dump on demand
  • get the core file
  • if you are using lldb version 3.8 (not needed if you are using lldb version 3.9 or later):
    • segment it with a script from llnode ./llnode/scripts/readelf2segments.py ./core > core.ranges
    • let lldb know where to look export LLNODE_RANGESFILE=core.ranges
  • run lldb -c core
  • use v8 commands to analyze

How to find stuff in a core dump

TBD

When should I use core dumps?

TBD