I am wondering if any one knows of a way to manage Xyratex disk shelves from CentOS (in particular CentOS 4).
More details: Some years ago I installed a NAS unit from Exanet which consists of 2 rebadged IBM x3650 head nodes and a couple of Xyratex disk shelves with a total of 96 TB of raw disk, connected by fibre channel. The operating system is based on CentOS 4.4, but is modified, and runs a proprietary file system. It has pretty good performance and I'm happy with it.
You may know that the company went bust some years ago, so I have been running this system unsupported. I am loath to give up the disk space though.
The two things I worry about are, the fibre channel connection and the external disk trays. Unfortunately, what I know about fibre channel is very little. They are emulex HBAs, and I guess there used to software called lputil, but this is no longer available at the Emulex site. It has been replaced by something called HBAnywhere, but I have not succeeded in building that on CentOS 4. Uprading the OS is out of the question I would think while maintaining the file system capability.
And the other question is how to manage the LUNs on the Xyratex's.
I'd be grateful for any info anyone has on either.
Tony Schreiner
On 08/16/2013 07:51 PM, Tony Schreiner wrote:
Have you tried to locate ex employees of that Company? Maybe you could hire some developer from that Company to help you out?
On 8/16/2013 10:51 AM, Tony Schreiner wrote:
Some years ago I installed a NAS unit from Exanet which consists of 2 rebadged IBM x3650 head nodes and a couple of Xyratex disk shelves with a total of 96 TB of raw disk, connected by fibre channel. The operating system is based on CentOS 4.4, but is modified, and runs a proprietary file system. It has pretty good performance and I'm happy with it.
I would consider dumping the head nodes, and connecting the FC arrays directly to the host HBA, and running them as fiber JBOD to a dedicated host, which could run whatever sort of mdraid, lvm, file system you want. or something like FreeNAS with ZFS and FreeBSD, then share stuff via SMB, NFS, etc.
On Aug 17, 2013, at 5:16 PM, John R Pierce wrote:
It may come to that eventually. Though I don't know why use say dump the head nodes, they are adequate servers (unless you mean overwrite them).
The main reason I'm not doing that now, is that the Exanet software provides coherent caching between the 2 nodes for enhanced redundancy and performance.
Additionally, I am stuck with the LUNs provides by the Xyratex until I am able to find a way to manage them.
Tony
-
John R Pierce -
Ljubomir Ljubojevic -
Tony Schreiner