Lustre discuss - Oct 2007 - Redundance with Lustre

Hi,

I''m consideranding to install Lustre in my job''s cluster.
Scenario is: 19 dedicated nodes, with 250 GB SATA disk (OST) and a 
server node (MDS) with RAID 1. My objetive is to create a fs with Lustre 
using free space in each node. I have readed the manual, but I have 
doubts about redundance. What happens if a disk fail (hardware failure)? 
Will the data in that disk be lost? Is there any kind of RAID5 o 6 
between the disk mounted in the nodes (distribuited RAID to prevent 
hardware failures)?

Thanks in avant.

On Oct 10, 2007  00:16 +0200, Jos? M. Mart?n wrote:
> I''m consideranding to install Lustre in my job''s cluster.
> Scenario is: 19 dedicated nodes, with 250 GB SATA disk (OST) and a 
> server node (MDS) with RAID 1. My objetive is to create a fs with Lustre 
> using free space in each node. I have readed the manual, but I have 
> doubts about redundance. What happens if a disk fail (hardware failure)? 
> Will the data in that disk be lost? Is there any kind of RAID5 o 6 
> between the disk mounted in the nodes (distribuited RAID to prevent 
> hardware failures)?
Currently there is no such feature.  We are working to implement "Server
Network Striping" to allow such lustre-level redundancy for systems that
do not have shared disks.

Cheers, Andreas
--
Andreas Dilger
Principal Software Engineer
Cluster File Systems, Inc.