On Tuesday 19 February 2008 16:38:24 Anna Jonna Armannsdottir wrote:
On þri, 2008-02-19 at 10:32 +0200, Buchan Milne wrote:
If you *really* *really* want to do this, it can be done. Run a bb
localhost 'hobbitdboard host=xxxx test=conn color=red' to check the
status,
if it's down, you can use 'net' from samba to reboot the server (.e.g
'net
rpc shutdown -r -S server -U user , with an account that has
sufficient
rights.
Hi Buchan,
this is an interesting and inspiring idea. It would probably be
unreliable in case the connection of the windows host is down.
Then the net command would not work.
Sure, and in my environment, I don't do this anyway. If I had to, I would
probably use 'ipmitool -H server chassis power reset' ...
Most modern servers have built-in watchdogs, so that if a certain
condition is met (generally frozen machine) the watchdog reboots
the machine.
Yes, I disable these features on many of our HP servers, as the servers reboot
for no really valid reason with this feature enabled. Since we use iLOs or
ILOMs for fencing (and not the NMI watchdog), I prefer not to have the
randomness in my environment.
It is a built in hardware/firmware thingie, that
has its own OS, and network interface.
Usually the watchdog itself is in the BIOS, not the SP/ILO/ILOM etc.
Then is the question of how to make such a watchdog listen to the
Hobbit monitor. Most of such firmware has built in SNMP but i do
not know if any of them would be able to catch a SNMP TRAP and
then reboot upon that.
In this scheme, the role of the Hobbit monitor would be to monitor
and send an alert to maybe a devmon or some other process that would
convert that alert into a SNMP trap and send it to the watchdog.
But, I'm not sure exactly what the OP wanted to do ... I merely posted an
example that should work for virtually any circumstance via the OS.
Regards,
Buchan