Environment
Sentinel 7
Situation
SUSE Linux Enterprise Server (kernel version
3.0.101 onwards) have stopped the automatic loading of watchdog
drivers and this system change has an impact on Sentinel high
availability implementations. After upgrading to a kernel version
that includes this change, the cluster services do not come up
automatically after the system boots because the watchdog
driver may not be auto loaded.
Resolution
All customers who have implemented high availability with SUSE
Linux Enterprise Server should contact their hardware vendor to
identify the recommended watchdog driver for their Sentinel server
and follow these steps:
Modify /etc/init.d/boot.local to include the following line. Then either reboot your machine or run the command once to load the driver immediately.
/sbin/modprobe -v --ignore-install "watchdog driver"
Cause
Recent kernel upgrades of SUSE Linux Enterprise
Server (kernel version 3.0.101 onwards) have stopped the automatic
loading of watchdog drivers. Previously, when sbd opened
/dev/watchdog (if no driver was loaded yet) the kernel would
auto load everything that could possibly serve as a watchdog driver
and stick to the first one that didn't fail to load. This did not
work properly. SUSE recommends that customers identify the watchdog
driver recommended by their hardware vendor. Then check whether it
has been loaded by doing the following:
lsmod | grep <watchdog driver>
If it auto loads, it should be present in the output. If it
isn't present, it should be added to the start script on the
operating system as directed in the resolution section.
Additional Information
The SUSE
Linux Enterprise Server team recommends the following site for more
information:
http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/drivers/watchdog/Kconfig.
You can look at the Makefile to map kernel configuration symbols to
driver names as needed.
Although it is generally
better to use hardware specific drivers on physical hardware, there
is also a universal watchdog driver called softdog
that might offer enough safety margin for your needs. NetIQ
recommends that you contact your hardware vendor for more
guidance.