I've been doing RAID for decades and this configuration makes no sense to me whatsoever. We have about a thousand Solaris systems in our environment in a variety of configurations and I've never seen anything like this. First, RAID-1 is true mirroring and if it's being done by the controller, the whole disks are being mirrored. Unless you're totally paranoid, or are detaching one of the mirrors for backups or DR, there would be no reason to mirror three drives. It may be that they're actually set up as a mirrored pair of two drives with a spare that the controller can use if one of the mirrored disks fails. But even that's unnecessary if the system is being properly monitored and a drive failure would be promptly noticed and corrected. Wasting a drive in this fashion flies in the face of your "budget limitations". The additional layering of RAID-5 makes no sense for several reasons. If the disks are already mirrored, ther's no reason for another layer of RAID on top of that. If you had several sets of mirrored disks, you might want to put a RAID-0 stripe across them for performance (as in RAID-10) but you wouldn't use RAID-5 and there's only one mirror here. Secondly, the slices you mention are on the same drive. It's very rare that a drive would fail in such a way that one slice would fail while another slice would be OK, so this doesn't provide any additional. Also, the use of two slices on the same drive would cause that drive to be extremely busy as it tries to read and write two parts of the RAID-5 group that are on the same spindle. Lastly, RAID-5 requires a minimum of three slices, so I don't see how this one was built with only two parts. All of this makes even less sense if the "spindle" from the OS point of view is already in RAID-1 at the controller level. Every time you access the RAID-5 group, all three actual spindles must go hyperactive trying to reach the slices of the group that are on the same spindle. If your description is accurate, I would have to conclude that whoever configured this server was not at all familiar with Sun metadevices and RAID concepts. If you have a chance to reconfigure/rebuild some day, there are many of ways to better utilize the disks in this system to improve performance and provide data security.

I've been doing RAID for decades and this configuration makes no sense to me whatsoever. We have about a thousand Solaris systems in our environment in a variety of configurations and I've never seen anything like this. First, RAID-1 is true mirroring and if it's being done by the controller, the whole disks are being mirrored. Unless you're totally paranoid, or are detaching one of the mirrors for backups or DR, there would be no reason to mirror three drives. It may be that they're actually set up as a mirrored pair of two drives with a spare that the controller can use if one of the mirrored disks fails. But even that's unnecessary if the system is being properly monitored and a drive failure would be promptly noticed and corrected. Wasting a drive in this fashion flies in the face of your "budget limitations". The additional layering of RAID-5 makes no sense for several reasons. If the disks are already mirrored, ther's no reason for another layer of RAID on top of that. If you had several sets of mirrored disks, you might want to put a RAID-0 stripe across them for performance (as in RAID-10) but you wouldn't use RAID-5 and there's only one mirror here. Secondly, the slices you mention are on the same drive. It's very rare that a drive would fail in such a way that one slice would fail while another slice would be OK, so this doesn't provide any additional. Also, the use of two slices on the same drive would cause that drive to be extremely busy as it tries to read and write two parts of the RAID-5 group that are on the same spindle. Lastly, RAID-5 requires a minimum of three slices, so I don't see how this one was built with only two parts. All of this makes even less sense if the "spindle" from the OS point of view is already in RAID-1 at the controller level. Every time you access the RAID-5 group, all three actual spindles must go hyperactive trying to reach the slices of the group that are on the same spindle. If your description is accurate, I would have to conclude that whoever configured this server was not at all familiar with Sun metadevices and RAID concepts. If you have a chance to reconfigure/rebuild some day, there are many of ways to better utilize the disks in this system to improve performance and provide data security.