I've now settled on OpenSolaris and using ZFS for the speed and loveliness of it all in comparison to LVM+XFS. But now I'm left with a dilema about how to configure the intital disk layout. As I'm still very much a newbie sysadmin, I was wondering what other more experienced sysadmins would do to configure the following:

I'm using comoddity hard drives, these are what I have to play with.

  • 2x 1TB
  • 2x 750Gb
  • 5x 500Gb
  • 2x 320Gb

The core OS is installed in a 60Gb SSD.

I have the following to begin with to build this new setup:

  • 2x 1TB
  • 3x 500Gb

Once the other linux server is decomissioned, I'd like to add:

  • 2x 750Gb
  • 3x 500Gb
  • 2x 320Gb

I need to have a redundant setup for about 500Gb (this is where all our source data will reside - emails, documents etc). Another 2-3Tb of FAST accessible data which has lots of video files (most are 3-4Gb in size, some are 1.4Gb MPEG and raw AVI).

What would be the suggestions? I'm still unsure if I can do a RAIDZ1 with the amount of drives I have and the idea of Storage Pools etc.

Here is our layout currently - in Linux:

  • /home/ (320Gb)
  • /media/Storage (500Gb)
  • /media/Scratch (~2.5TB = 2x750 + 2x500)
  • /media/Musical (500Gb)

The musical folder is not required to be redundant, neither is Scratch as its just housing our production videos that are copied to DVDs most of the time.


Well, the amount of disks of the same size isn't ideal (it would be better to have more disks of the same size). Having said that I would create two pools:

  1. A RAIDZ2 pool consisting of 4x500GB: vdev of 3 drives for RAIDZ2 (=> 500GB) and one spare disk:
    zpool create datapool raidz2 500gb_1 500gb_2 500gb_3 spare 500gb_4

  2. A pool consisting of three mirror vdevs: 2x1TB, 2x750GB and 2x320GB (=> ~2TB):
    zpool create scrpool mirror 1tb_1 1tb_2 mirror 750gb_1 750gb_2 mirror 320gb_1 320gb_2

You would have one 500GB drive left. If you add it to the first pool (you need to do this right from the beginning!) you could double its size to 1TB.


  • RAIDZ2 survives a double disk failure.
  • You have a spare device for the RAIDZ2. I.e. full redundancy can immediately be restored once there is a drive failure.
  • You still have some redundancy in the scratch pool because of the mirrors. I.e. one drive in each mirror vdev can fail and your video files are still okay. But unfortunately you have no spare disk.
  • If you REALLY don't want redundancy for scratch (I don't recommend this) you COULD add all the six drives to the second pool without mirroring to get a 4 TB pool. But it would probably die soon because one single drive failure is enough to break it.
  • Hi, Thats VERY much for your reply! That was a great bit of information, unfortunately I can't vote you up yet for it :( Can I ask you what a good disk setup is? Should I ask to buy a bunch of 500Gbs instead? The place I'm working for is very constrained budget wise and its hard to get the head honcho to agree to buy new HDDs when we already have an abundant amount. – Leah Calum Oct 12 '09 at 9:03
  • You don't have to. You can start with what you have although the missing spare disk in the scrpool is not optimal. By the way: Make sure you haven enough bandwidth for all disks and don't oversubscribe your controller(s). – knweiss Oct 12 '09 at 22:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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