[Pacemaker] IPMI Stonith

Adrian Chapela achapela.rexistros at gmail.com
Tue Jan 20 15:14:00 UTC 2009


Dejan Muhamedagic escribió:
> Hi,
>
> On Mon, Jan 19, 2009 at 08:00:50PM +0100, Adrian Chapela wrote:
>   
>> Hello!
>>
>> I am thinking again on the stonith. IPMI could be an option, but is it 
>> really a good option for production ? What happen if the network break 
>> down ?
>>     
>
> You need to provide different paths for the heartbeat and
> connection to the ipmi devices.
>
> Also, make your power supply resilient: if a node loses power,
> the ipmi device will be dead too, and your cluster will wait
> forever for the fencing to succeed.
>   
Ok, but what about a kernel panic ? when the OS hungs up the other node 
can't know what is happening on the node that is hunged up. I am 
thinking on this situation. Of course the two node will have RAID, 
redundant network (with bonding) and redundant power supplies but there 
is an option to have a split brain with a node hunged up, what do you 
think about this ?
>   
>> Is totally resolved the split brain ? Have you tested with HP 
>> Proliant Servers ?
>>     
>
> I think that people use ipmi with proliants.
>
> Thanks,
>
> Dejan
>
>   
>> Thank you!
>>
>> _______________________________________________
>> Pacemaker mailing list
>> Pacemaker at oss.clusterlabs.org
>> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>>     
>
> _______________________________________________
> Pacemaker mailing list
> Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
>   





More information about the Pacemaker mailing list