Take the 2-minute tour ×
Server Fault is a question and answer site for system and network administrators. It's 100% free, no registration required.

Irecently asked a question about the best way to apply 20 2TB drives in a ZFS pool. I have come up with the following setup (thanks to many helpful suggestions) and would love opinions and suggestions about what I may have missed.

I have two SATA controllers (controller1 & controller2) that have the following disks attached:

  • controller
    • disk1
    • disk3
    • disk5
    • disk7
    • spare1
  • controller2
    • disk2
    • disk4
    • disk6
    • disk8
    • spare2

My thought was to create the pool using mirror pairs of disks, one disk each from each controller; ie:

  • mirror1: disk1 & disk2
  • mirror2: disk3 & disk4, etc

The two spare drives would give me some redundancy and also having the mirrors made up of disks from each controller should give me some redundancy as well.

Have I missed anything? Ii feel like I can grow the pool with little fuss by adding mirrored pairs and also more redundancy by added spare pairs.

I feel like this would give me a good balance of speed and redundacy.

Thoughts welcome and appreciated.

share|improve this question

3 Answers 3

That works just fine. Yes, you definitely want your mirrored disks to be distributed across controllers. That way, you could sustain a controller failure. You haven't missed anything there.

share|improve this answer

My only fear with this setup is the amount of bandwidth that the controllers can handle seeing as you will have X amount of mirrors across the controller.

share|improve this answer

I'm was running same setup in production. IOPS bases its your best possible setup (not taking any Log, ZIL, ARC and L2ARC in account). The bad side of this is that you will face data loss if both disks fail in one 2-way mirror. That sounds like a rare case, but I had to face this 3 weeks ago. It's not that rare if you start thinking about it. It's quite common that all the disks you've are from the same production batch. So if one fails some others will follow. When one of the disks in the 2-way mirror fails a spare one drops in and resilvering will start. When it starts the still working disk will get quite some load because it has to copy all its data to the new spare disk. In my case the other disk did go down too. So you could also go with multiple 3-way mirrors or multiple raidz2 or even raidz3. I'm running on 3-way mirrors after last incident. Of course this is all less important if you have perfect backups. Good luck!

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.