On Mon, 14 Mar 2011, Louis Guillaume wrote:
I have two 2TB drives. They need to be mirrored and cut up in to
several chunks that can be exported via iSCSI and/or NFS. So my first
inclination is to create 1 large wedge on each disk, RAID-1 them
together with RaidFrame and then split up the raid using wedges.
Now this is not that different, I guess, than disklabels on raidframe
on disklabels. But it does bring up a few questions for me...
o Is there any benefit to creating wedges on raid0 vs. disk-
labeling it?
o Consider that the "partitions" on raid0 will be exported via
iSCSI. What "type" would I use in dkctl raid0 addwedge?
o Is there any advantage to sharing up the wedges (or partitions)
vs. creating a filesystem and sharing up sparse files? I can't
imagine the latter being better for performance!
o Any other pitfalls wrt wedges that I should be concerned about?
If you are using istgt or iscsi_target, these will tend to use large
(sparse if you want) files as the backing storage. You can export
multiple LUNs all from one filesystem, so depending on your NFS
requirements, there's no reason to chop up your backing store.