[PVE-User] Questions about fencing setup

[Stefan Sänger]

Honestly, I am a bit disappointed by your answer. I was hoping that you 
could point to some documentation or explanation. I probably missed some 
aspects of man pages right away, so I figured asking here woul be of 
help. Anyway...

Am 14.04.2012 07:53, schrieb Dietmar Maurer:
>> First, what does<method name="1">  in clusternode fence config mean?
>
> This is just an identifier for the fence method

So, it is a method identifier. Great. Let's have a closer look:

Each device used for fencing is configured in <fencedevice> - section of 
/etc/pve/cluster.conf. The parameters there are communication parameters 
- so with this information the cluster knows *how* to talk to the fence 
device.

In <clusternode><fence> there is a setting that creates an identifier 
for the fence method used.

But - I don't see any hint on what fence action is configured. Logical 
assumption is, that it uses the default action as stated in man page, 
and that is given as "reboot".

And if I wanted to use another action like power off I would have to 
configure that in fencedevice-section.

Am I right?


>> Will it power off the server in case of fencing action is required?
>
> Usually a reboot is better.

Probably. But that was not my question.


>> If so - I assume that the iDRAC access will be dead as well and I have to
>> power on the server again manually, right?
>
> Hopefully not - you should test that.

Usually I just try for myself and ask the silly questions later. In this 
case the servers are several hours of travelling away, so It would be 
kind of uncomfortable to just issue a command that could possibly lead 
to a power off.

>> Second - and probably more important - what is 192.168.7.180 here:
>> <rm>
>>       <service autostart="1" exclusive="0" name="TestIP" recovery="relocate">
>>         <ip address="192.168.7.180"/>
>>       </service>
>> </rm>
>>
>> Is this some kind of ping node to see if the network is still alive?
>
> This is an example how to configure a failover IP address.
>
> Simply do not use if you do not understand it.

Well - what do you think why I asked?
Of course the file listings given in proxmox wiki are supposed to be 
examples. But you may admit that given documentation is still very rough.

After looking through the man pages and the wiki pages again, I should 
change my guess; probably this adress will be up as long as the cluster 
is up. So it is probably used to access cluster administration 
regardless of cluster state, e.g. usually it is up on first cluster 
node; if that node fails it will be up on another cluster node.


>> And at last - let's assume the node died because of hardware failure,
>> e.g. defunct mainboard or even better - defunct PDU in redundant PSU.
>>
>> After fixing the hardware issues, how to bring the node back online again?
>
> Press the power on button?

And it will automatically be unfenced again? And perform a failback?



TIA,
Stefan