[Linux-cluster] RE: < fecing with out any hardware? >

Pool Lee, Mr <14117614@sun.ac.za> 14117614 at sun.ac.za
Mon May 1 19:45:55 UTC 2006


What about software fencing? Is it really nesasary to be hardware!

Is there a difference between lutre/cfs, the product that sun uses, and gfs?

I'm planning to do mostly numerical work with the cluster and thus I would like all the machines to be able to 
retrieve data, as if it was local on the machine. NFS is very limited in this regard because we intend on using vast arrays of matrices, that can be up to 1-2 Gig.

I was hoping to implement GFS since all the machines are already setup, without the hardware fencing though.

Kind Regards

He who has a why to live can bear with almost any how.
Friedrich Nietzsche

-----Original Message-----
From: linux-cluster-bounces at redhat.com on behalf of Bowie Bailey
Sent: Mon 2006/05/01 05:42 PM
To: linux clustering
Subject: [Linux-cluster] RE: < fecing with out any hardware? >
Pool Lee, Mr <14117614 at sun.ac.za> wrote:
> I was wondering if it would be a good idea to fence without any type
> of hardware, besides the pc's. 
> At the moment I have about 6 machines that I want to have a gfs on
> these 6 machines but due to budget constraints I cant afford to by
> hardware. How is it possible to fence without the use of hardware,
> besides manual fencing.   
> These machines are your basic desktop pc's, each has a 80gig HD and a
> 3Ghz P4 processor. They are all connected by a 1 Gigabit switch. 
> Would if be possible to use GFS or is there any other variant. They
> all run FC5( Fedora Core 5 ). I need some sort of GFS, because we
> intend on setting up a mysql clustering system as well.  
> Any ideas would be greatly appreciated.

You need to have some sort of fencing.  You can use manual fencing,
but it doesn't work well with production systems.  What happens is
that any problem in the cluster causes everything to come to a dead
stop and wait for you to fix the problem and then let the cluster know
it's ok to continue operation.

The only real option for a production system is some sort of hardware
or software that allows for the cluster to fence misbehaving nodes on
it's own.  The cheapest is a power switch.


Linux-cluster mailing list
Linux-cluster at redhat.com

-------------- next part --------------
A non-text attachment was scrubbed...
Name: winmail.dat
Type: application/ms-tnef
Size: 3942 bytes
Desc: not available
URL: <http://listman.redhat.com/archives/linux-cluster/attachments/20060501/a1b1dcc5/attachment.bin>

More information about the Linux-cluster mailing list