[Linux-cluster] After APC switch was called with fence_node - server stays shut down - what can be modified?

[Marc - A. Dahlhaus [ Administration | Westermann GmbH ]]

Am Mittwoch, den 16.09.2009, 11:14 -0500 schrieb Alan A:
> I am using HP Proliant DL385s - new ones with Dual Quad Core AMD
> Opteron(tm) Processor 2384 and 32G of memory. I am using APC Power
> Switches for fencing. When I issue fence_node clusternode1 - sometimes
> it works like a charm and sometimes server stays down, like it was
> shutdown, although power on the APC switches was reinstated. I can see
> the green light on the APC Switches but if the node stays shutdown
> there is no green light on the Power Supplies. What could be wrong,
> what should I check?

Hello Alan,

this looks like a bios configuration issue. Take look if you can find
anything "power loss" related in bios and change it from "off" to "last
state" or "on". I don't recall what it was called exaclty on the
proliant bios or where you can find that option in bios setup...

Hope this helps.

Marc

OT: You should also make usage of the iLO management devices in this
boxes, they are rock solid fencing devices and react faster than the
apc-power switches (we use DL360s and APC PDUs as fallback fencing
here)...