Grow as I Go

[MtK]

Hey,
I have an Server that I want to set as a backup server (mostly to store backups). and I have 6 x 2TB for it.

The problem is not all the disks are free at this moment and I only have 1.
another 2 are still in the old backup server, so I need a new backup server to be able to move the backups to it.
another 3 HDDs will be get free at a later stage.

What and how can/should I build my server so I could add the HDDs as they get free?

MtK

 

[survive]

Hi MtK,

Assuming you want to use zfs & want the drives to all end up in a redundant (raidz, raidz2) pool you can't.

zfs doesn't do disk-by-disk expansion.

-Will

 

[MtK]

Do I have any other options?
maybe 3 now and 3 later...?

 

[survive]

Hi MtK,

Give this a look:

http://forums.freenas.org/threads/slideshow-explaining-vdev-zpool-zil-and-l2arc-for-noobs.7775/

-Will

 

[Caesar]

if you have 3 disk now you can create a single 3-disk vdev in Raidz and then expand your pool by adding second vdev of 3-disk also Raidz or backup your data, destroy the 3-disk vdev and create a 6-disk vdev Raidz2

FYI - A single 6 disk vdev in Raidz2 is better than two 3-disk vdevs in Raidz

 

[MtK]

@Will, I already read this before I posted.
it doesn't really talk about growth, or degraded raids.

@Caesar, I thought 2 3-vdevs is better then 1 6-vdevs in terms of IOPS, because it is mirrored/stripped.



Anyway, I found 4 x 750GB, so with the 1 x 2TB I already have I thought about creating a degraded RAIDZ of 6-vdevs (5 + a missing 1).
Then when all the 2TB disks are freed I'll just upgrade the POOL.

Sounds good?

 

[Caesar]

@Caesar, I thought 2 3-vdevs is better then 1 6-vdevs in terms of IOPS, because it is mirrored/stripped.[/qoute]

I took backup server to mean that you preferred redundancy over performance so in terms of redundancy it is better to have 1 vdev with 6 drives than 2 vdevs with 3 drives.

Anyway, I found 4 x 750GB, so with the 1 x 2TB I already have I thought about creating a degraded RAIDZ of 6-vdevs (5 + a missing 1).
Then when all the 2TB disks are freed I'll just upgrade the POOL.

I do not believe that you can create a degraded pool. I think you have to have all of your disk when the pool is created. If you don't care about redundancy why not create 5 single disk pools until you get all of your 2TB disk then you can create the 6 disk pool and move your data over.

 

[fracai]

I haven't tested this, or even really thought it out, but I think it might be possible to go the "degraded 6 disk RAIDZ2" route.

1) Create a RAIDZ2 pool using 4 x 750 GB, 1 x 2 TB, and a temporary file of some insignificant size (say, 1 MB) (possibly by making the FreeNAS boot volume writeable). You'll have a pool that is severely limited in size (6 x 1 MB, minus parity, etc).

2) Detach the temporary file. You'll have a degraded pool (essentially a 5 x 1 MB RAIDZ1).

3) Detach the 2 TB drive. You'll have a degraded pool (essentially a degraded 5 x 1 MB RAIDZ1).

4) Replace the temporary file with the 2 TB drive, you may have to wipe it on another computer first. If the "autoexpand" property was set to "on" (it is by default?) you should end up with a degraded 6 x 750 GB RAIDZ2 (essentially a 5 x 750 GB RAIDZ1).

You can then replace the drives (including the missing device) with the 2 TB drives as they're available.

Again, I wouldn't trust this with valuable data, but it'd be an interesting experiment regardless.

 

[titan_rw]

You can create a sparse file with dd and seek, then mdconfig the sparse file to a /dev/mdX device. Then use the /dev/mdX device during a manual zpool creation (cli). Then fail out the /dev/mdX device(s). Destory the md devices, and delete the sparse file. No need to deal with autoexpand, as the pool is created with devices of the right size to begin with.

Definitely an advanced way of migrating data. I've used it before, and it does work, but you're doing a lot of the work the gui does for you through the command line.

 

[MtK]

Thanks @fracai & @titan_rw, that's the general idea.

but the good news are, I found another 250GB drive, so with 1 x 250GB + 4 x 750GB + 1 x 2TB, I can actually create a (somehow) full RAIDZ2 of 6 x 250GB.
then simply replace drives as I go, until I have a full 6 x 2TB.

This way, data is transferred to a more secure storage and actually big enough for it...

right?

 

[paleoN]

I haven't tested this, or even really thought it out, but I think it might be possible to go the "degraded 6 disk RAIDZ2" route.

You can create a sparse file with dd and seek, then mdconfig the sparse file to a /dev/mdX device.

Just an FYI, this method is posted in the How-To section. ;)

 

[MtK]

Just an FYI, this method is posted in the How-To section. ;)

yes, that's why I thought about this option.
but since I have 6 drives I won't be needing it... :)