[Pvfs2-users] PVFS and DRBD

[Michael O'Sullivan]

Hi everyone,

I am very new to storage networking. My research group (NDSG, 
www.ndsg.net.nz) is using optimisation to automatically design storage 
networks, but we need to get the right software, redundancy, clustering, 
etc to make our designs work correctly.

Our current storage system is very simple. Two servers (HP Proliants 
currently running Windows Server 2003 although it looks like we will ned 
to change to Linux, we have support for CentOS) that access disks on two 
storage devices using an ethernet core-edge network. The storage devices 
are dual-core motherboards (running CentOS) with two 1 TB sata disks 
each. We want the storage devices to be share the workload, to provide 
redundancy for each other and to appear as a single (clustered) storage 
system.

What I would like to do is as follows:

* Set up one DRBD node with disk 1 on (storage) device 1 as the primary 
disk and disk 2 on device 2 as the secondary disk.
* Set up another DRBD node with disk 1 on device 2 as the primary disk 
and disk 2 on device 1 as the secondary disk.
* Cluster the two DRBD nodes together using PVFS or something similar.

This means that if device 1 fails then device 2 will still contain all 
the data for the system to keep functioning. If both devices are up and 
running then device 1 will provide DRBD node 1 and device 2 will provide 
DRBD node 2 and share the load of the storage system. I ws hoping 
someone would be able to let me know if this is possible and, if so, 
provide some pointers on how to get started.

Any help is greatly appreciated. Thanks, Mike