On Tue, 03 Mar 2009 20:16:33 +0100
Joerg Barfurth <[email protected]> wrote:

> Jens Langner schrieb:
> 
> > If you can instruct me how to create such a core dump in a
> > running/productive SRSS environment, or if you can tell me where the
> > core dumps of utauthd are normally stored I can have a look and try
> > to send them over to you.
> > 
> 
> On Solaris, you can use utcoreadm to configure a place (and naming 
> pattern) for collecting coredumps of all crashing processes.
> 
> You can use something like
> 
> # mkdir /var/cores
> # coreadm -g /var/cores/gcore_%f_%u_%g.%t -e global
> 
> utauthd core dumps show up as gcore_java_0_0.<timestamp> files.
> 
> Be sure to monitor /var/cores, so that it doesn't grow too big. I 
> usually make /var/cores a separate zfs dataset with a quota on my 
> production machines.
> 
> If you are on Linux, I am not sure of the best way to achieve the
> same effect ...
> 

Most Linux distributions use ulimit to prevent the writing of corefiles 
in case of a segfault etc:

# ulimit -c 
0

The ulimit value is inherited by child processes, so you have to add

 ulimit -c unlimited

somewhere in the startup script of utauthd.

The place and name of core files is controlled by kernel parameters:

# mkdir /var/cores
# chmod 777 /var/cores
# echo 1 > /proc/sys/kernel/core_uses_pid
# echo /var/cores/core_%e_%t > /proc/sys/kernel/core_pattern

will create cores with names 
/var/cores/core_<filename of executable>_<timestamp>.<pid>

Meik

-- 
Meik Hellmund
Mathematisches Institut, Uni Leipzig
e-mail: [email protected]
http://www.math.uni-leipzig.de/~hellmund
_______________________________________________
SunRay-Users mailing list
[email protected]
http://www.filibeto.org/mailman/listinfo/sunray-users

Reply via email to