Upgrading with a ZFS dataset in place

[thany]

I just attempted an upgrade from 6.11.5 to 6.12.1, and had to downgrade after discovering that my ZFS datasets were gone. It has my VMs and dockers on it, so they are pretty much critical to my setup, and I can't really spend too much time trying to repair it. Therefore I has opted for a downgrade (for now) to gather up some insight on what could go wrong, and how to better prepare for this upgrade.

 

So in short, what should I do to retain my ZFS datasets?

 

Is there a migration guide for the ZFS plugin to native unRAID ZFS? I'm sure someone must've written up something, because I'm not the only one using this plugin, and it's clearly not a 0-step in-place upgrade. I have to do *something* to keep my dataset intact, question is, what.

 

Can someone please provide a kick in the right direction?

[Iker]

Hi there, you can import your existing pools without too much trouble, the process is very well outlined in the 6.12 release notes https://docs.unraid.net/unraid-os/release-notes/6.12.0/#zfs-pools; however, please be aware of some limitations:

• Not all pool topologies are supported
• Please be carefull about mountpoints, as those may change from whatever location you may have to /mnt/<pool_name>
• Autotrim and compression can be configured for the entire pool
• First level datasets will be imported as shares, you have to change some settings in the Shares section and select the most appropiate config for primary an secondary storage (See https://docs.unraid.net/unraid-os/release-notes/6.12.0/#share-storage-conceptual-change)

If you find any erros or have any issues please report.

Edited June 22, 2023 by Iker

[thany]

Thank you. If I'm not mistaken, it says I just create a new pool and are allowed to assign existing ZFS devices to it, which will then "import" a previously created pool. If that works, that's excellent, but I've made a data backup just to be on the safe side, just in case it decides to clear all my data as part of creating the pool

 

The `zpool status` command reports my pool as `raidz1-0` so I think I should be fine, since `raidz1` appears to be supported. Fingers crossed.

[thany]

It worked perfectly!

 

The only thing that wasn't quite clear is that the array has to be stopped in order to modify anything around ZFS. I wasn't used to that, but found out quickly enough.