Solaris

Now this doesnt look right to me. All of these disks are 100Gb LUNS so total zpool size is 300Gb.

Am I right in saying that this zpool consists of two disks plus one more disk that is 6 way mirrored?

So a bit pointless because only one of the three is mirrored (and 6 way is a bit of overkill anyway)

Code:

zpool status data01
  pool: data01
 state: ONLINE
  scan: none requested
config:

        NAME        STATE     READ WRITE CKSUM
        data01      ONLINE       0     0     0
          c1d2      ONLINE       0     0     0
          c1d3      ONLINE       0     0     0
          mirror-6  ONLINE       0     0     0
            c1d12   ONLINE       0     0     0
            c1d13   ONLINE       0     0     0
            c1d14   ONLINE       0     0     0
            c1d15   ONLINE       0     0     0
            c1d16   ONLINE       0     0     0
            c1d17   ONLINE       0     0     0

A single disk failure of c1d1 or c1d2 in that zpool will result in complete data loss.
Zpool consists of those above and a 6 way mirror vdev called mirror-6 between rest.

What worries me as well is that those are LUNs, and are protected on storage level (correct me if i'm wrong)
Adding additional raid layer is not advised or desired, and will surely impact performance of both storage system and host.

If those are not raid protected outside the system, and performance is required, one should configure using 8 x 100 gb as 4 top level vdevs configured as mirrored pairs of two disks.

This will give 400 GB zpool which consisting of 8 disks which can tolerate upto 4 devices failing, as long as the failed disks are not of the same top level vdev.

If both disks fail from one mirror (vdev), you will lose data from zpool.

One should introduce a hot spare disk in above configuration (the 9th disk) to minimize such probability to minimum.
Of course, a instant death of 2 disks (at the exact time and space) will cause zpool data loss.

This is why you backup for those statistically improbable but possible conditions.


Hope that helps.
Regards
Peasant.

Yes I thought that was the case. Its now been rebuilt as 4 top level mirror paired disks.

In answer to you're question though - no at the moment its not raid in the storage appliance although this is coming soon and, as you say, we will halt the mirroring at the host end.

Hot spare disks - Tell me more. What is command to do this? I assume it uses this disk if any of the 8 disks fails at any point?

Yes, will hot spare added to pool and if autoreplace option is on in the zpool the following will happen.

1. FMA agent detects a fault and replaces the failed device with spare.
2. When the failed disk is replaced and resilvered, hot spare disk is detached and added back to the pool as hot spare.

Other then that, if the option is not defined, the administrator will need to do option 2 by hand.

This is documented quite well.

Regards
Peasant.