Tertiary mirror is separate but synchronized with RAID set so writes are performed synchronously
• To create a point-in-time copy:
– Applications must be quiesced for a few seconds – Pending transactions are flushed from buffers
– Mirroring is stopped and split mirror can now be accessed by other processes
– Application is allowed to continue
Primary RAID Split
Mirror
Application server
Split Mirror Backup server, etc
All in same array
We have our primary volume, our primary volume could be RAID 1+0, it could be RAID 5 or RAID 6.
We write to our primary logical unit, our split mirror will take that data and move it over. When we update, we write to our primary storage and if these are in a paired state the update also goes across to our mirror.
When we’re ready, we quiesce and we flush, we’ve now got our consistent usable copy of data so we can now do our split. This now becomes 100% copy of our data on a separate set of spindles (a separate RAID group) and now when changes come in we update our primary but we don’t update the split mirror because this is our point in time copy.
It can help protect us against corruption but it also protects us against primary volume failure. Because if this failed we have a full copy of our data at a given point in time. It is still in-system replication.