Looking for input on a possible sysklogd bug

[kubicle]

I'm Looking for some input on a possible syslog bug on feisty. I'd like to know whether this suspected 'bug' affects other users as well as my machine.

/etc/cron.daily/sysklogd script causes syslogging to halt when it's run (The script is run as a daily cron job to rotate logs)

In the end, the script should restart the syslog daemon, but it fails to do so, effectively killing logging.

I was able to 'quick-fix' it by editing the script:
/etc/init.d/sysklogd reload-or-restart > /dev/null
to
/etc/init.d/sysklogd restart > /dev/null
(forcing a restart instead of reloading)

Before I dive deeper into it, I'd like to know whether the 'bug' affects all machines or is there just something wrong on my system

Thanks

(P.S. I noticed there was an old similar dapper bug report on launchpad)

[sky]

Re: Looking for input on a possible sysklogd bug

Hi kubicle,

On my system I don't see any trouble, my feisty is started from 13 days and it has 7 occurences of /var/log/syslog.*
syslog is running.
How do you see this trouble and from when ?

Cheers

[kubicle]

Re: Looking for input on a possible sysklogd bug

Hi kubicle,

On my system I don't see any trouble, my feisty is started from 13 days and it has 7 occurences of /var/log/syslog.*
syslog is running.
How do you see this trouble and from when ?

Cheers

Thanks for the info

There are basically two things I've modified on my system that may affect it. firstly I've set syslog to listen to remote log messages from my router/firewall, and secondly I've set log outputs to be shown on VT5 (router/firewall) and VT6 (systemlog).

Bug symptoms:
I noticed yesterday that the logs had fallen silent and that the last messages were from cron daily jobs. And after manual troubleshooting I noticed that after running the sysklog cron job the /var/log/syslog (the active log file) was recreated as an empty file and no further messages are logged into it (stays as an 0kb file)

As I mentioned, editing the script fixes it on my system, but I'll debug further when I find the inspiration (my primary canditate is the 'reload' part of /etc/init.d/sysklogd init-script, especially the HUP signal sent by it, but that's just a gut feeling...I'll probably edit the script to run more verbosely to see where the culprit lies)

[kubicle]

Re: Looking for input on a possible sysklogd bug

I think I've narrowed it down to the 'reload' case of /etc/init.d/sysklogd

here's the relevant section:

case "$1" in
start)
log_begin_msg "Starting system log daemon..."
create_xconsole
start-stop-daemon --start --quiet --pidfile $pidfile --name syslogd --startas $binpath -- $SYSLOGD
log_end_msg $?
;;
stop)
log_begin_msg "Stopping system log daemon..."
start-stop-daemon --stop --quiet --pidfile $pidfile --name syslogd
log_end_msg $?
;;
reload|force-reload)
log_begin_msg "Reloading system log daemon..."
start-stop-daemon --stop --quiet --signal 1 --pidfile $pidfile --name syslogd
log_end_msg $?
;;
restart)
log_begin_msg "Restarting system log daemon..."
start-stop-daemon --stop --retry 5 --quiet --pidfile $pidfile --name syslogd
start-stop-daemon --start --quiet --pidfile $pidfile --name syslogd --startas $binpath -- $SYSLOGD
log_end_msg $?
;;
reload-or-restart)
if running
then
$0 reload
else
$0 start
fi
;;
*)
log_success_msg "Usage: /etc/init.d/sysklogd {start|stop|reload|restart|force-reload|reload-or-restart}"
exit 1
esac

Here's the same from the edgy script (6.10):

restart|force-reload|reload-or-restart|reload)
log_begin_msg "Restarting system log..."
start-stop-daemon --stop --quiet --exec $binpath --pidfile $pidfile
sleep 1
start-stop-daemon --start --quiet --exec $binpath -- $SYSLOGD
log_end_msg $?
;;

Is the daemon supposed to wake up by itself after the SIGHUP signal, as the feisty reload case only invokes 'start-stop-daemon --stop' (as opposed to 'start-stop-daemon --stop' & 'start-stop-daemon --start') on edgy...or should the feisty script use stop-and-start as well?

EDIT: Oh, thanks to Birdy for pointing out the differences in the edgy script

[sky]

Re: Looking for input on a possible sysklogd bug

Hi,

It's really strange because my scripts are exactly the same as yours, and it's working... Sorry I can't help more.

Good luck

[penguin.ch]

Re: Looking for input on a possible sysklogd bug

I think I've narrowed it down to the 'reload' case of /etc/init.d/sysklogd

To that I agree on the grounds following:

ROOT@HAL # ps -u root | grep 'sys' -
7213 ? 00:00:00 syslogd
ROOT@HAL # /etc/init.d/sysklogd reload
* Restarting system log... [ ok ]
ROOT@HAL # ps -u root | grep 'sys' -
7582 ? 00:00:00 syslogd

ROOT@KUB # ps -u root | grep 'sys' -
4308 ? 00:00:00 syslogd
ROOT@KUB # /etc/init.d/sysklogd reload
* Restarting system log... [ ok ]
ROOT@KUB # ps -u root | grep 'sys' -
4308 ? 00:00:00 syslogd

That one got me hooked (Kubicle, you must not feed an iguana ...).
So I changed the relevant passage of the "new" sysklogd script to:

log_begin_msg "Reloading system log daemon..."
start-stop-daemon --stop --quiet --pidfile $pidfile --name syslogd
start-stop-daemon --start --quiet --exec $binpath -- $SYSLOGD
log_end_msg $?

ROOT@KUB # ps -u root | grep 'sys' -
5437 ? 00:00:00 syslogd
ROOT@KUB # /etc/init.d/sysklogd reload
* Restarting system log... [ ok ]
ROOT@KUB # ps -u root | grep 'sys' -
5638 ? 00:00:00 syslogd

Go figure ...