Debugging random crashes

It's very rare for a CentOS box to crash, but recently one of our boxes was crashing at random times.

In order to work out what was happening, I used the "crash" software that comes pre-installed nowadays.

To start with, you need to install the kernel-debuginfo package (yum install kernel-debuginfo).
If it's not available, go into /etc/yum.repos.d/CentOS-Debuginfo.repo (or stream equivalent) and set enabled=1.

You then need to go to your crash directory (typically /var/crash), and load up the crash report with the following command:

crash /usr/lib/debug/usr/lib/modules/$(uname -r)/vmlinux vmcore

It will then tell you what happened and with what piece of software.

In our case we saw:

PANIC: "BUG: unable to handle kernel paging request at ffffd6963eab17a8"
PID: 76524
COMMAND: "chrome"

Indicating an error in Google Chrome (which we'd just updated).

If no crashes are stored, make sure kdump is installed and running, and check settings @ /etc/kdump.conf



Who's amazing? Sarah is.

Want to get in touch?