[Linux-cluster] fence problem on rhel5

[Chris Harms]

I do not have cluster storage, e.g. SAN, GFS, NFS.  Using 2 DRBD 
patitions for postgres so I will evaluate this option carefully.  If it 
can take care of its own split brain recovery and avoid FS corruption, I 
may chance the lack of fencing as a nonfunctional cluster is pretty bad.


Lon Hohberger wrote:
> On Thu, Jul 12, 2007 at 01:57:57PM -0500, Chris Harms wrote:
>   
>> What are the implications of this?  i.e. , if there is node failure and 
>> fenced is not running, does the cluster skip fencing and assume services?
>>     
>
> That's what's supposed to happen.
>
> Obviously if you have shared storage, you can expect data corruption too
> ^^;;
>
>