ZFS setup question (Newbie)

[bunk3m]

Hi.

I've been trying to get Nas4free going for a while and want to use ZFS. I've been reading about ZFS for a week now and am still somewhat confused.

I've installed Nas4Free in a VM (ESXi 5.1) and direct connect two 3TB hard drives via M1015 flashed in IT mode. I realize this isn't the preferred method to run Nas4Free but I'm trying to consolidate our home IT infrastructure on 1 box.

I can see the drives and I was able to format them ZFS.

I want to set the two drives to mirror since I can't do Raidz. I would like to get there eventually by adding more drives when I have more money. (Is this possible to change to Raidz later?)

I can't get the zpool going.

I've set the two drives as a virtual devices (Disk1 & Disk2 for lack of better names). Then I thought I have to set the zpool, give it a name (FileStorage1) select the two vdevices and select mirror. When I click Accept it gives an Error 1.

Is this the right way to go about this?

Thanks in advance.

[siftu]

Firstly you can not change from a mirror to a RAIDZ at a later date, you can add more virtual devices to the pool i.e. another mirror pair or even a RAIDZ (but try to keep a pool with same raid levels). If you need to change you have to destroy your pool and recreate it.

code 1 just means thata process returned a failed exit status, you can generally see what went wrong at the bottom of the log. (Diagnostics|log).

[bunk3m]

Thanks for the tip. I didn't realize there was a log screen available.

From the system log I get these messages when I create (1) Pool:Virtual Device with 2 3TB drives and set to mirror because of 2 disks only, (2) Pool:Management - try to create a pool selecting the vdevice with 2 drives. Error happens when I try to apply the create pool command.

Code: Select all

Aug 17 14:28:27 	nas4free 	root: gnop: Provider da1.nop already exists.
Aug 17 14:28:27 	nas4free 	root: gnop create -S 4096 /dev/da1

da1 is actually the first disk attached.

My apologies if this is so obvious but I'm still totally confused.

[bunk3m]

I think the problem might have been that I should have set up the two drives as 1 udev with mirror and the zpool with the 1 udev as this appears to have worked. I thought after reading the documentation that udev was like a disk, pool was like multiple disks with redundancy. Apparently I got it wrong.

So I have 2 3TB disks as 1 udev of 3TB because of the mirror. The pool is of the 1 udev of 3TB. So I believe (assuming you don't correct my misunderstanding) that I have some redundancy because of the mirroring of the two disks.

I'll add an external backup to complete the cycle before I start adding more disks. I presume this makes the most sense in a home environment??

Thanks everyone for your help. Much appreciated!

[siftu]

I think the problem might have been that I should have set up the two drives as 1 udev with mirror and the zpool with the 1 udev as this appears to have worked. I thought after reading the documentation that udev was like a disk, pool was like multiple disks with redundancy. Apparently I got it wrong.

So I have 2 3TB disks as 1 udev of 3TB because of the mirror. The pool is of the 1 udev of 3TB. So I believe (assuming you don't correct my misunderstanding) that I have some redundancy because of the mirroring of the two disks.

I'll add an external backup to complete the cycle before I start adding more disks. I presume this makes the most sense in a home environment??

Thanks everyone for your help. Much appreciated!

You are correct.. You set up redundancy at the vdev level and keep adding those to a pool.. the pool will stripe among it's vdev members

[fsbruva]

the pool will stripe among it's vdev members

Not correct. The pool might stripe across it's members. It might not. ZFS will balance writes across the vdevs with differing i/o speeds as it sees fit. Labeling it as "striping" implies some equality of data across the vdevs that simply cannot be assumed.

[siftu]

the pool will stripe among it's vdev members

Not correct. The pool might stripe across it's members. It might not. ZFS will balance writes across the vdevs with differing i/o speeds as it sees fit. Labeling it as "striping" implies some equality of data across the vdevs that simply cannot be assumed.

Yes you are right, technically SUN/Oracle call it "dynamic striping"

Multiple RAID-Z groups can be dynamically striped to form a larger storage pool.

http://docs.oracle.com/cd/E19253-01/819-5461/gazdd/