My notes, things I have learned, and documentation I have created.

To enable core dumps for an xinetd process, I needed to install abrt. On my minimum install of RHEL6, I needed to install the following to get core dump files created:

1:dbus
abrt
abrt-addon-ccpp
abrt-addon-kerneloops
abrt-addon-python
abrt-cli
abrt-libs
abrt-plugin-logger
abrt-plugin-rhtsupport
abrt-plugin-sosreport
ConsoleKit
ConsoleKit-libs
eggdbus
elfutils
elfutils-libs
libtar
polkit
sos
xmlrpc-c
xmlrpc-c-client
xz
yum-utils

I must still be missing something, because I get the following when I try to list crashes.

abrt-cli –list
dbus error: Failed to connect to socket /var/run/dbus/system_bus_socket: No such file or directory
error requesting DBus name com.redhat.abrt, possible reasons: abrt run by non-root; dbus config is incorrect; or dbus daemon needs to be restarted to reload dbus config

And I am not able to start the daemon (abrtd). I get the following when I try to start the daemon:

Starting abrt daemon: abrtd: Failed to start: timeout waiting for child

It looks they are probably related, and there are a few bugzilla reports on the issue, but no clear solution. I just needed to get the core dumps created, and I have that for now.

I am fairly certain that the mods to sysctl.conf do not do anything as I have never gotten a core file in /tmp as a result of this change. I don’t know if the limits settings are needed or not, I left them in place.

The key to getting core dumps in RHEL6 is abrtd. In my case I had to:

* make sure that the entries for abrt were in /etc/passwd and /etc/group, eg:
abrt:x:499:499::/etc/abrt:/sbin/nologin (for passwd)
abrt:x:499: (for /etc/group)

* make sure that abrtd would start. In my case I had to turn it on with /sbin/chkconfig. It also would not start correctly via the /etc/init.d/abrtd script until I added the passwd and group entries for the user/group. Check your syslogs for “abrt” information about startup errors.

* get the /etc/abrt/abrt.conf file configured correctly. In my case I set:

OpenGPGCheck = no
ProcessUnpackaged = yes
MaxCrashReportsSize = 0 (for unlimited cores)
comment out all entries for ActionsAndReporters

Once I did these steps, I could fire up a process, do a “kill -3” on the process from another window, and have it say “Abort (core dumped)”. Another scan of the syslogs told me that the core stuff got written to /var/spool/abrt. A directory ccpp-[numbers] got created there with a file “coredump” inside. At that point I could do “gdb [path to executable] coredump” and get going on debugging. Finally!

What idiot at Redhat made this the default behavior for application core dumps? While abrtd is a good idea, don’t force it on us to do standard unix core debugging.

If anybody finds the idiots name in Redhat who made the decision to F up a simple thing like core file creation; let u know. I’d like kick his arse… What cooking mess …

I’m having this error when creating database in red hat 6 Linux ;
An unexpected error has been detected by Hotspot Virtual Machine :
SIGSEGV (0*b) at PC=0*9f40945c, pid=6141, tid=3079419584
Java VM:Java Hotspot(TM) client VM (1.5.0_17-b02 mixed mode)
problematic frame:
c {libnnZ11.so+0*3c45c}
An error report file with more information is saved as hs_err_pid6141.log
Aborted(core dumped)
please help me out .