Setup to take ganesha coredumps

Redhat RHEL7.x

Automatic Bug Reporting Tool (ABRT) is the preferred method to take application coredumps on Redhat Enterprise Linux systems. The following steps can be used to configure an RHEL 7.x system to take Ganesha coredumps:

1. Install the abrt-cli RPM if not already installed:

   yum install abrt-cli
   

2. IBM ganesha package are not signed, so you need to configure ABRT to take coredumps from executables belonging to unsigned packages:

   Set OpenGPGCheck=no in the /etc/abrt/abrt-action-save-package-data.conf file.
   

3. Allow coredump to be unlimited in size (limited by the space available in /var/spool/abrt filesystem):

   Set MaxCrashReportsSize = 0 in the /etc/abrt/abrt.conf file.
   

4. Start (or restart) the abrt daemon service:

   systemctl restart abrtd
   

5. Start (or restart) the abort-ccpp service:

   systemctl restart abrt-ccpp
   

6. A core dump might not be generated for code areas where the Ganesha process changed its credentials. To take coredump from all paths, fs.suid_dumpable should be set to 2:

   Insert the following entry into the /etc/sysctl.conf file:
   fs.suid_dumpable = 2
   
   Run the following to make the above setting effective:
   sysctl -p
   

7. Verify that kernel.core_pattern is set to abrt-hook-ccpp (something like |/usr/libexec/abrt-hook-ccpp %s %c %p %u %g %t e %P %I %h and fs.suid_dumpable is set to 2:

   # sysctl kernel.core_pattern
   kernel.core_pattern = |/usr/libexec/abrt-hook-ccpp %s %c %p %u %g %t e %P %I %h
   
   # sysctl fs.suid_dumpable
   fs.suid_dumpable = 2
   

8. Verify that the system actually takes Ganesha coredumps when it crashes! Here we send abort signal to Ganesha process to intentionally crash it, please note that the Ganesha daemon will be terminated after this and the generated coredump should be deleted as it is useless:

   # kill -SIGABRT <Ganesh-daemon-PID>
   

For additional details, see the Documentation about ABRT-specific configuration.