Hi, I posted in the Solaris install forum as well about the fileserver I''m building for media files but wanted to ask more specific questions about zfs here. The setup is 8x500GB SATAII drives to start and down the road another 4x750 SATAII drives, the machine will mostly be doing reads and streaming data over GigaE. -I''m under the impression that ZFS+(ZFS2) is similar to RAID6, so for the initial 8x500GB two drives would be sucked into parity so I''d have a 3TB volume with the ability to lose two discs, no? -I can''t add discs to the RAIDz volume so in the future for the 4x750 I''d just do a new RAIDz on them (one disc is for parity) and add it to the ZFS group. -If I do that it won''t write data across both RAID volumes will it? -If I want more control over what files go on which array should I keep them as distinct volumes? Thanks, Sam This message posted from opensolaris.org
On 2/5/08, Sam <sam at smugmug.com> wrote:> > Hi, > I posted in the Solaris install forum as well about the fileserver I''m > building for media files but wanted to ask more specific questions about zfs > here. The setup is 8x500GB SATAII drives to start and down the road another > 4x750 SATAII drives, the machine will mostly be doing reads and streaming > data over GigaE. > > -I''m under the impression that ZFS+(ZFS2) is similar to RAID6, so for the > initial 8x500GB two drives would be sucked into parity so I''d have a 3TB > volume with the ability to lose two discs, no?Yes, zfs2 eats two disks for parity. -I can''t add discs to the RAIDz volume so in the future for the 4x750 I''d> just do a new RAIDz on them (one disc is for parity) and add it to the ZFS > group.No, you can''t add disks to a raid group right now. You wouldn''t want to anyways, as if you put the 750''s into the raid group with the 500''s, only 500GB would be used on them. -If I do that it won''t write data across both RAID volumes will it? Yes, it will. You have one pool with multiple raid groups in it, and it does a pseudo "raid-0" across both pools. -If I want more control over what files go on which array should I keep them> as distinct volumes? >Correct. You won''t really have any control over which files go to which raid group if you have them all in the same pool. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://mail.opensolaris.org/pipermail/zfs-discuss/attachments/20080205/2c940cae/attachment.html>
> -I''m under the impression that ZFS+(ZFS2) is similar > to RAID6, so for the initial 8x500GB two drives would > be sucked into parity so I''d have a 3TB volume with > the ability to lose two discs, no?"RAIDZ2" is the term you''re looking for; and yes, you''d wind up with 3 TB of usable space.> -I can''t add discs to the RAIDz volume so in the > future for the 4x750 I''d just do a new RAIDz on them > (one disc is for parity) and add it to the ZFS > group.Hmmm. That''s not a good idea, because you''ll lose the ability to lose [any] two disks and survive. If you wanted to pool all of the disks, and you plan to use RAIDZ on one group of disks, you might as well use RAIDZ on both, IMO.> -If I do that it won''t write data across both RAID > volumes will it?Yes, if you add them both to the same zpool.> -If I want more control over what files go on which > array should I keep them as distinct volumes?Yes. You should create two zpools, one for each array. Filesystems are tied to a single pool, so at that point you have full control (at the filesystem level) of which files go on which array. This message posted from opensolaris.org