Suppose I were to split the 2TB SSD into two partitions. Let's suppose that this results in the two partitions being called disk1s1 and disk1s2. We also have the two 960GB SSD devices in the SATA bays, and these are assigned disk2 and disk3. Would it be possible / advisable to form a zpool that pairs one partition with one of each of the smaller disks, like so (skipping the finer details of normalization, case-sensitivity, ashift, etc):
- Code: Select all
% zpool create TANK mirror disk1s1 disk2 mirror disk1s2 disk3
So, this would pair up a partition of the 2TB SSD with one 960GB (resulting in a 960GB mirror VDEV), and a second partition of the 2TB SSD with the other 960GB SSD. If one of the 960GB SSDs fails, there is redundant data in the corresponding 2TB partition, and if the 2TB SSD fails it takes out both partitions but the two 960GB SSDs still have redundant data to carry the load. Does anyone see a problem with this scheme? Obviously it uses up 3 SATA bays, which is not ideal, but it seems like I would gain security. It all hinges on whether it is a good idea to pair up a partition with a disk; I've put zpools on partitions in the past on my laptop, but never used them in mirrors.
Thank you for your advice.