Re: watchdog failure at same time every week

Mike F wrote:
>
> we are testing Oracle 9i RAC.Every sunday morning @ 0102 - 0104 the
> watchdog fails to communicate and shuts down one node and then the
> whole database. The machines then reboot. This happens without fail.
> It happened on RedHat 7.1 and it happens on SUSE 7.2. Our SA are not
> away of any crontab jobs runing at that time. Any help would be
> appreciated.
>
> NODE 2
> wdd.log
> UTC: Wed Oct 24 19:00:07 GMT 2001 (746103)
> wddProcRegisterPacket: info: registered client
> name = /tmp/.watchdog/cl_sock_788_15373,

I guess there is indeed a cron job running once a day/week... which removes old files in /tmp. We had a similar problem on our SuSE boxes: The script /etc/cron.daily/aaa_base_clean_tmp removed the unix domain socket file of the listener in /var/tmp/.oracle. Very tricky!

On SuSE you can reconfigure this very easily via /etc/rc.config (TMP_DIRS_TO_CLEAR, OWNER_TO_KEEP_IN_TMP). On RedHat there must be a similar mechanism.

Hope it helps

Uwe

-- 
Uwe Schneider       | Telefon +49 7244 / 609504
Karlsdorfer Str. 31 | Mail    uwe_at_richard-schneider.de
DE-76356 Weingarten | http://www.richard-schneider.de/uwe
Linux - OS al dente!