I have a need to created pool that only concatenates the LUNS assigned to it. The default for a pool is stripe and other possibilities are mirror, raidz and raidz2. Is there any way I can create concat pools. Main reason is that the underlying LUNs are already striping and we do not want to stripe in ZFS *and* in the storage cabinet (internal politics). To my knowledge the only other possibility (if ZFS can''t do it) is using SVM. Greetings, Peter -- This message posted from opensolaris.org
On Wed, Jan 28, 2009 at 07:37, Peter van Gemert <opensolaris at petervg.nl> wrote:> Is there any way I can create concat pools.Not that I''m aware of. However, pools that are not redundant at the zpool level (i.e., mirror or raidz{,2}) are prone to becoming irrevocably faulted; creating non-redundant pools, even on "intelligent" storage arrays, is thus not recommended.> Main reason is that the underlying LUNs are already striping and we do not want to stripe in ZFS *and* in the storage cabinet (internal politics).What reasons are there for not doing so? Perhaps if we know more about the situation we can suggest alternative configurations. Will
On Wed, 28 Jan 2009, Peter van Gemert wrote:> I have a need to created pool that only concatenates the LUNS > assigned to it. The default for a pool is stripe and other > possibilities are mirror, raidz and raidz2.Zfs does "concatenate" vdevs, and load-shares the writes across vdevs. If each vdev is one disk or one LUN, then you have concatenation and not "striping". Bob =====================================Bob Friesenhahn bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/ GraphicsMagick Maintainer, http://www.GraphicsMagick.org/
Peter van Gemert wrote:> I have a need to created pool that only concatenates the LUNS assigned to it. The default for a pool is stripe and other possibilities are mirror, raidz and raidz2. > > Is there any way I can create concat pools. Main reason is that the underlying LUNs are already striping and we do not want to stripe in ZFS *and* in the storage cabinet (internal politics). >Nobody has provided a convincing use case to justify concats, so they don''t exist. In the bad old days, logical volume managers did concats because they could not dynamically stripe. ZFS does dynamic striping (!= RAID-0) so it doesn''t have the capacity issues caused by RAID-0. -- richard> To my knowledge the only other possibility (if ZFS can''t do it) is using SVM. > > Greetings, > Peter >