On Wed, 2006-04-05 at 09:49 +0200, Rune V. Sjøen wrote:
2) enable core files, by changing the core file limit using "ulimit -c unlimited", and making sure syslog-ng runs in a writable directory (check /proc/<pid>/cwd and check if it is really writable)
/proc/<pid>/cwd points to / syslog-ng runs as root. How can I change the cwd to be absolutely sure it runs in a writable environment ?
if syslog-ng runs as root it should dump core to / if it is not read-only.
3) wait for syslog-ng to dump core, and then send me the syslog-ng binary and the core file in private.
as far as I can see there are no core files created when it dies, or should I say 'disappear'.
Hmmm. An strace dump but one that follows the syslog-ng process as well should help which kind of signal makes syslog-ng to exit. Try to attach strace to syslog-ng once it has started by using: strace -o strace.dump -f -s 256 -p <pid of syslog-ng process> If it is some kind of heap corruption the only way to solve the problem is to have the core file. (and one core is not necessarily enough) Thanks for your help, it is appreciated. -- Bazsi