New user with questions...

[crimsondr]

Hi all,

I am considering using FreeNAS to setup a file server. I currently have a file server running Ubuntu with 6x1TB drives in RAID5. I am interested in creating a new server with either 6x4TB or 6x3TB drives. From my research it seems that zfs would be the way to go.

My questions:
1. Do zfs snapshots serve the same purpose as rsnapshot?
2. Is it safe to grow/shrink a zfs filesystem? Is the data vulnerable during a grow/shrink operation? I would be interested in maybe adding drives to expand the zfs volume in the future without having to copy off all the data and re-create the volume.
3. My current Ubuntu install also runs virtualbox and plex media server. Are there any problems running these on a FreeNAS install?

Thanks.

 

[Ericloewe]

Hi all,

I am considering using FreeNAS to setup a file server. I currently have a file server running Ubuntu with 6x1TB drives in RAID5. I am interested in creating a new server with either 6x4TB or 6x3TB drives. From my research it seems that zfs would be the way to go.

My questions:
1. Do zfs snapshots serve the same purpose as rsnapshot?
2. Is it safe to grow/shrink a zfs filesystem? Is the data vulnerable during a grow/shrink operation? I would be interested in maybe adding drives to expand the zfs volume in the future without having to copy off all the data and re-create the volume.
3. My current Ubuntu install also runs virtualbox and plex media server. Are there any problems running these on a FreeNAS install?

Thanks.

1. No idea what an rsnapshot is. Care to explain?
2. You can't shrink a ZFS pool. Expanding is easy, but requires proper planning (you can't do one drive at a time).
   
3. They're hackier, but many people use them.

For details on ZFS, I recommend Cyberjock's guide (link is in my sig).

 

[crimsondr]

rsnapshot is a package of scripts that uses rsync and hard links to make full snapshots of your data. You can have full daily snapshots but it only uses the space of 1 full backup + incrementals.

I just read that while expanding a zfs pool the system is a degraded state, so if anything bad happens your data is gone.

I will look at Cyberjock's guide. Thanks.

 

[Tywin]

I just read that while expanding a zfs pool the system is a degraded state, so if anything bad happens your data is gone.

This is not necessarily an issue, because you should probably be using RAIDZ2 anyway. With RAIDZ2, you will still have redundancy while resilvering each drive member, so you will be single fault tolerant.

 

[danb35]

I just read that while expanding a zfs pool the system is a degraded state, so if anything bad happens your data is gone.

In addition to @Tywin's answer, this is going to depend heavily on exactly how you're expanding the pool and how your pool is configured. If you have a RAIDZ1 pool (similar to RAID 5, and not recommended), and you're expanding it by pulling each drive and replacing it with a larger one, one at a time, yes, the pool will be degraded while you're doing that. But there are other ways of configuring your pool to begin with, and other ways of expanding it:

• If you have chassis space and SATA ports to spare, you can "replace" a disk in a RAIDZ1 array without removing it--that is, you can add the new disk, tell FreeNAS to replace the old disk with the new disk, and wait for the resilver to finish. The old disk won't be kicked out of the array until the resilver is finished and your data is intact on the new disk, so your array won't be degraded.
• If you have a RAIDZ2 array (similar to RAID 6, and recommended over RAIDZ1), when you remove one disk to replace it with a larger one, the array will be degraded, but it will still have redundancy.
• You can always expand your pool by adding another redundant array. If your pool consists of a two-disk mirror, you can add another two-disk mirror. If you have a three-disk RAIDZ1, you can add a six-disk RAIDZ2. There's no requirement that the vdevs match, and there isn't even a requirement that the new vdev have any redundancy--but if it doesn't, and if it fails, you lose your entire pool.

It's a point worth repeating: Although it's possible to expand your pool by adding single disks, it's strongly discouraged. One article I read refers to this practice by its technical name, "hating your data." Data is striped across all vdevs in a zpool, so if any vdev fails (i.e., if your newly-added single disk fails), your entire pool will be lost.

 

[crimsondr]

Can you replace a disk without removing it in any raidz configuration? The example was for raidz1 but can you do the same with raidz2 or raidz3?

I just finished reading Cyberjack's guide. It was very informative.

All the warnings makes zfs appear very delicate but he may have been just trying to emphasize the point.

So if any vdev dies, your zpool is gone. So it's best to have each vdev with at least raidz2 or better. Bad memory (non-ecc) can cause the zpool to become corrupt during scrubs (not sure what that is at the moment). Improper shutdowns can also corrupt the zpool. And usually any zpool corruption leads to all loss of data. Sound about right?

My current ubuntu file server has had at least one drive die in the past 6 years. I also have that hooked up to a UPS and it shuts down automatically when a power failure is detected.

 

[crimsondr]

@danb35, looking at your signature you have running exactly what I want. PMS, VirtualBox and transmission. The only other thing I would need is flexget for automatic downloads and transmission.

 

[crimsondr]

This is not necessarily an issue, because you should probably be using RAIDZ2 anyway. With RAIDZ2, you will still have redundancy while resilvering each drive member, so you will be single fault tolerant.

That is true. I would most likely use RAIDz2.

 

[Tywin]

• you have chassis space and SATA ports to spare, you can "replace" a disk in a RAIDZ1 array without removing it--that is, you can add the new disk, tell FreeNAS to replace the old disk with the new disk, and wait for the resilver to finish. The old disk won't be kicked out of the array until the resilver is finished and your data is intact on the new disk, so your array won't be degraded.

That is awesome, I didn't realize you could do this :) Can you do multiple disks at a time? Say, all of them?