FreeBSD 12 to FreeNAS Migration

[silkie]

I've been using FreeBSD since about version 4 for my home NAS and I'm comfortable there but I spun up a FreeNAS install and liked what I saw. So much so I'm moving to it. Why reinvent the wheel right?

The trouble with good technology is you can build and forget because it always just keeps working! And that is where I've come unstuck... basically I'm saying I'm a bit long in the tooth and I've not read about your new fangled ways of doing things! I'm worse than a n00b in many respects. :)

So, I have a question about data migration really and some tips about how best I might slice up my new drives. When I built my FreeBSD NAS ZFS wasn't even a thing. It's evolved over the years to a respectable ZFS zRAID2 pool backup up to UFS volumes, it's been running like this for many years!!! From when I initially built my ZFS setup I can see people are a lot more confident with it as a technology these days so you keen youngens I've hoping can help with a little advise.

My plan is a simple one.

I have 11 new disks for my new zRAID2 array 10 for data 1 as hot spare arriving this weekend.

My current setup is 9 disks, 8 in zRAID2 and a hot spare. The pool is backed up to 4 UFS volumes using and nightly rsync.

I plan to physically remove the current array (pop it in a cupboard in case something goes wrong! Build the new disk into a new array from scratch within the FreeNAS WebGUI.

Here are the questions: -

How do I recreate my datasets from the UFS backups? FreeNAS system volume seems to be read only as fstab updates are not persistent... would there be any trouble forcing a RW mount and editing fstab to mount the UFS volumes just for the sake of copying the backup data back into the pool? I can remove it later once I have a plan for backups later.

Once my data is on the new array what are the best ways to backup to these disks in the new world order of FreeNAS?? Would that be to bring the disks into the FreeNAS server as 4 single disk ZFS volumes and setup ZFS exports and imports? My rsync scripts will probably still work with a little tweaking but if there are better ways to do this I'm open to learning.

How best might I carve up the 10 disks. I'm thinking single vDEV 1x10 zRAID2 seems OK, not fast but most reliable. It's just a home server for media and doc storage. I run a small vsan VMware installation, some iSCSI might be nice but that could be built into a new SSD based zpool I have physical space for. I run 2 IBM 1015 HBAs at the mo.

Any and all advice would be gratefully received.

 

[dlavigne]

Did you decide upon a solution for this?

 

[silkie]

I'm making progress. The new raidz2 array is running + spare and has my data on from using the UFS disk import feature. Regular scrub schedule is testing now, snapshots will be enable there after. I need to read about local backups re-purposing those UFS disks once I'm happy I'm rolling forward with this. I'm slowly building up all my services that my old FreeBSD NAS used to provide, some I can, some I can't. Overall, I'm happy.

 

[Chris Moore]

You might find some tips in this guide if you still have questions:

fester112:intro [danb35's Wiki]

www.familybrown.org

 

[silkie]

I am now ready to start the backup of the RAID based datasets to my local single disks. Fester hasn't got there yet with those handy walk throughs. Is it the "Replication Tasks" section where I should be looking? Would ZFS replication be more efficient than rsync jobs?

BTW: OneDrive sync is very nice you clever bunch!!

 

[Chris Moore]

danb35 or jgreco might have some tips they would point you to. I have used ZFS sned | ZFS receive to copy data pool to pool and if you search the forum, you will certainly find some directions on that. If you are bringing data from a UFS partition, you would likely be fairly well servd with a regular rsync or copy command.

 

[Elliot Dierksen]

If you are bringing data from a UFS partition, you would likely be fairly well servd with a regular rsync or copy command.

The way I have done when moving from one set of drives to another residing in the same system is cd /source-directory ; find . -print | cpio -pvdm /destination-directory . This a particularly useful from a non-ZFS source. I agree with @Chris Moore that ZFS send | ZFS recv would the way to go when source and destination are both zfs.

 

[jgreco]

I am now ready to start the backup of the RAID based datasets to my local single disks. Fester hasn't got there yet with those handy walk throughs. Is it the "Replication Tasks" section where I should be looking? Would ZFS replication be more efficient than rsync jobs?

BTW: OneDrive sync is very nice you clever bunch!!

rsync can be incredibly efficient depending on the specifics of how you deploy it. When there are no changes, it can take just a handful of seconds to check for differences on 40,000 files totalling 4GB.

 

[silkie]

I have loaded the data from my old UFS backup volumes (there are 4) onto the new zRAID pool. I've now created a new single zfs disk pool within the WebGUI using one of the re-purposed UFS disks. Looking at the zfs send | receive WebGUI pages and that of the rsync pages they are focus on sending data to remote systems. I've settled on local rysnc commands entered into the cron job WebGUI pages. First "backup" is running right now, I will repeat on the other 3 disks in due course. I hope I'm on the right tracks. A local backup within the same tin isn't as safe as an offsite but it's better than nowt! These cron'd rsyncs will run nightly.

I have these two 240GB IronWolf SSDs sat here spare, ordered before I realised I could USB boot. Anyone got any ideas how I might use them? I have HBA ports for them I just can't think of what to do with them. I ran a nice ZIL/L2ARC on a pair of old 60GB San-disk SSDs in FreeBSD, now it seems everyone wants a SLOG what ever that is. It would give them a reason for being! I'm thinking a separate HBA would be a good idea. SSD's could saturate my old HBA buses that would share the same HBAs as the main pool. More pondering!

Thanks for the help so far.