some ZFS questions

[Paul Kraus]

On Aug 7, 2014, at 4:16, Scott Bennett <bennett at sdf.org> wrote:

>     One thing I ran across was the following from the zpool(8) man page.
> 
>  "For pools to be portable, you must give the zpool command whole
>  disks, not just slices, so that ZFS can label the disks with portable
>  EFI labels. Otherwise, disk drivers on platforms of different endian-
>  ness will not recognize the disks.”

Here is where understanding *why* helps you make better decisions. In order for a zpool to be importable (is that even a word?) on a given system, the underlying OS must be able to read the parse the partition information. Since the *only* partition system that *all* the OSes that support ZFS understand is EFI, you need to use that if you want to have a truly portable zpool. I *think* you can manually create the EFI and just put zpools in the various partitions, but I am not positive.

If you only care about moving zpools between FreeBSD systems, and KNOW that you will NEVER want to import this pool on, for example, a Solaris system, then you can use any partition system FreeBSD understands. I have used GPT and moved zpools around between FreeBSD systems without a problem.

The critical part is that the OS must understand the partitioning and be able to present to the ZFS kernel module the data from within the partitions. Each partition that contains a zpool top level vdev will include 4 copies of the ZFS header, 2 at the start of the device and 2 at the end (note that I am saying device here, that could be an entire disk or a partition or even a file acting as a block device). So, if the ZFS code can read all 4 labels from the device, it can import this vdev, if it can import a sufficient number of vdevs that make up a zpool, it can import the zpool.

> If I have one raidzN comprising .eli partitions and another raidzN comprising
> a set of unencrypted partitions on those same drives, will I be able to
> export both raidzN pools from a 9-STABLE system and then import them
> into, say, a 10-STABLE system on a different Intel amd64 machine?

I am not familiar with FreeBSD .eli partitions, I assume at some layer below the block device the OS is encrypting the data here. If that is a valid assumption, then I’ll take a stab at answering :-) If the OS can read the data from the device and find the ZFS labels it can import that vdev (and with enough vdevs it can import the zpool).

>  By your
> answer to question 1), it would seem that I need to have two raidzN pools,
> although there might be a number of benefits to having both encrypted and
> unencrypted file systems allocated inside a single pool were that an option.

Oracle has done work on ZFS to permit native encryption within ZFS. I do not know off the top of my head if that is on a zpool or dataset basis. I expect that it is probably on a dataset basis, so you could have both encrypted and unencrypted datasets within one zpool (just as we can have both compressed and uncompressed datasets).

--
Paul Kraus
paul at kraus-haus.org