vdev setup questions

[jin]

Hey Guys,

Have been thinking of build a freenas server and have a few questions about different zpool setups.

Option 1:
4 2disk mirrored vdev
protect against 4 disk failures 1 from each vdev

Option 2:
2 4disk raidz2 vdev
protect against 4 disk failures 2 from each vdev

Option 3:
2 3disk raidz1 vdev
protect against 2 disk failures 1 from each vdev

Assuming 4TB drives all options should provide ~16TB usable space. My question is how do these differ during rebuilding in case a drive fails.

First off though is my understanding of vdev correct or did i mess that up?

Does mirroring have the least impact chance of failure when rebuilding
To upgrade i replace a disk in a vdev let it rebalance and repeat for all drives in a vdev. For raidz is the upgrade also done by recalculating parity and are there chances of failure during this time.

Thanks

 

[toadman]

It's a bit hard to read your questions (some are statements). I'll give it a shot. :)

The setup with the most redundancy will suffer catastrophic failure least often. That is option 2, it can survive any failure of two drives across the pool. The other options can trash the pool with the right combination of two drives failing.

Your understanding of vdevs is correct.

No, mirroring does not have the least chance of failure when rebuilding. See paragraph 1.

I'm not sure what you mean by "upgrade". If you mean growing the size of the vdev, then yes, you replace the disks with larger ones and resilver. When all disks are larger the pool will autoexpand. For RAIDZ, yes, parity is calculated by necessity. Anytime you resilver, mirror or parity, another drive in the system could fail. The ability to withstand additional failures depends on the pool topology.

The advantage of Option 1 is IOPS performance. You would have 4 vdevs vs 2. So the system will stripe writes across the 4 giving you double the IOPS. (Important if you're hosting VMs as an example.)

 

[Mryusef1]

I'm looking for some clarity. I have a pool setup with 4 4TB WD blue hard drives setup in a Raid2Z. If I understand correctly I can loose two drives and the pool with continue to work. The pool is showing up at 6.8 TB. When I had just two 4 hard drives stripped it showed up as 7.64 TB. Why has changed from stripped to raid2z cause me to loose space? Does ZFS use part of the drives like a swap file like in Linux? I have used Linux for a long time but I am new to FreeBSD.