Rebuild questions

[Cthulhu]

I've been happily using 9.2.1.5 in a 4x3TB RAIDZ1 configuration for some time now.

I'm needing to add more space, and intend to do the following:

* Add 4 3TB disks
* Reconfigure as 8x3TB RAIDZ2 or RAIDZ3
* Upgrade to the latest version.

Assuming I've got two backups of my current ZFS volume (rsync copies) and I want to keep my existing jails/plugins (2 instances of Plex and one of ownCloud), what's the best procedure to follow? (Obviously, I have to recreate the ZFS pool.)

 

[SweetAndLow]

I'm in the same boat. Using 9.2 and went to change layout. I was thinking of just rebuilding my jails from scratch and only backing up some jail meta data. Other options include zfs replications and just rsync the jails dataset.

 

[Robert Trevellyan]

I've got two backups of my current ZFS volume (rsync copies)

Any reason you can't replace one of those rsync copies with a replicated snapshot?

 

[Cthulhu]

Any reason you can't replace one of those rsync copies with a replicated snapshot?

I don't have access to another server to replicate the snapshot to - unless I'm missing some implication in your question.

I've got just under 4TB of data to back up which coincidentally just barely fits on an external USB 4TB drive. I have two of them at my disposal, and nothing else sufficiently large to hold all that data.

Tentatively, the plan looks something like this:

1. Back up the contents of the zfs poll to external drives (twice).
2. Take the server down.
3. Install four more drives.
4. Destroy the zfs pool and recreate as RAIDZ3.
5. (Restore the backed up data and restore jails/plugins.)
6. Upgrade to 9.3.

Restoring the data is no big deal, but I do want to make sure that my plugins retain their configuration and state.

 

[Robert Trevellyan]

unless I'm missing some implication in your question.

You can zfs send a snapshot to a file (using the command-line) or to a 2nd pool on an attached device (using the GUI or the CLI), it doesn't have to go to another server.

I do want to make sure that my plugins retain their configuration and state.

I think this will be easier using a snapshot.

Your plan for having two backups is wise.

 

[Cthulhu]

You can zfs send a snapshot to a file (using the command-line) or to a 2nd pool on an attached device (using the GUI or the CLI), it doesn't have to go to another server.

I think this will be easier using a snapshot.

Your plan for having two backups is wise.

Thanks, Robert - I knew I had to be missing something. :D

I'll look into zfs send/receive. I'm a little bit uncomfortable with the idea of using completely unfamiliar means to do a bare iron restore so I'm definitely going to use rsync for one of the backups just in case.

I'd take three backups if I had the space to do so.

 

[Cthulhu]

I need to move the existing motherboard, disks, etc to the new case. I won't be adding the new disks yet.

Although I'm going to make every effort to ensure the disks are going to be reconnected to the same SATA ports, is the array integrity going to be compromised if I don't?

 

[Ericloewe]

Although I'm going to make every effort to ensure the disks are going to be reconnected to the same SATA ports

Don't bother. It's absolutely irrelevant.

Even Intel's crummy fakeRAID is smart enough to be port-independent (and even system-independent).

 

[Ben Bliss]

If you have another desktop you could install freenas to a USB stick on said PC and then hook up the USB external drive and create a pool out of it. Then you could do a zfs send/receive.

 

[Cthulhu]

Finally finished "zfs send" on my snapshot - 27 hours to transfer to an external USB disk (USB sucks!).

Moved all the hardware to the new chassis, now I just have to wait for the new HBA to arrive to do the rest of the work. Everything went off without a hitch so far.

I'm really tempted to spin up an array on the new disks and test the restore before I destroy the old pool. I should at least hook up the drive and cp the files to /dev/null to make sure there's no read errors.

 

[Cthulhu]

The last couple of terabytes of data is being restored to the newly created 8x3TB RAIDZ2.

Thanks for the all the help, everyone.