This guide explains how to create a new, single disk ZFS pool under Linux. It does not cover more advanced ZFS configurations such as creating a mirror or a RAIDZ pool.
Note that you cannot take advantage of some of ZFS' best features when using a single disk ZFS pool, such as self-healing when performing zpool scrubs. You can however still use features such as snapshots so it is still worthwhile using ZFS even with single disk configurations as is often the case with removable and external drives.
This depends upon your Linux distro. Under Ubuntu you would run:
$ sudo apt install zfsutils-linux gdisk
To install the required software.
We need to know the device name of the disk we want to create our ZFS pool on so to do that we'll use the lsblk command. Attach the drive you wish to format then run lsblk in the terminal. The device name will usually be something like sdb or sdc etc. You should be able to identify the correct device by its size. If you have multiple drives of the same size attached, compare the output of lsblk with and without your drive attached.
Lets assume lsblk showed that sdb is the device I want to create my pool on. We will first remove all existing partitions on sdb so be VERY careful when running this next command because if you get the device name wrong you could end up formatting your system drive or any other attached drive and potentially lose ALL data on the drive so make ABSOLUTELY sure the device name is correct before running these commands! It is recommended you detach all of the drives that you do not need to have attached to your system before running this:
$ sudo sgdisk --zap-all /dev/sdb
You must replace /dev/sdb with /dev/sdc or whichever device lsblk tells you the disk is.
We will now create a Solaris ZFS partition on the same drive:
$ sudo sgdisk --new=1:0:0 --typecode=1:BF00 /dev/sdb
There are two ways you can create your pool when creating it under Linux. If you use the zpool create defaults, it will create a pool that can only be imported with read and write permissions on another sufficiently recent ZFS on Linux machine. If you want your pool to be readable and writable under FreeBSD, NetBSD, Solaris and Windows as well as Linux then a few extra ZFS options must be disabled when creating the pool.
If you are certain you will only ever want to import your pool (mount your ZFS disk) on Linux machines then you can run:
$ sudo zpool create -f -d -m none -o ashift=12 -O atime=off -o feature@lz4_compress=enabled backup01 /dev/sdb1
Replace backup01 with the name of your ZFS pool and replace /dev/sdb with the correct device. The pool name is the name you will use with zpool import to import (mount) the ZFS pool/disk.
If you don't mind losing some ZFS-on-Linux specific features and would prefer to create a multi-platform friendly ZFS pool instead, run:
$ sudo zpool create -f -d -m none -o ashift=12 -o feature@lz4_compress=enabled \
-o feature@multi_vdev_crash_dump=disabled -o feature@large_dnode=disabled \
-o feature@sha512=disabled -o feature@skein=disabled -o feature@edonr=disabled \
-o feature@userobj_accounting=disabled backup01 /dev/sdb
Again, replace backup01 with the name of your ZFS pool and /dev/sdb with the device with the empty Solaris partition to use for the pool. Pools created with this command should play nicely with FreeBSD versions 12.x and earlier. It shouldn't be necessary to disable features when sharing pools with FreeBSD 13.x and later because by that time it should've switched to using OpenZFS and as such will have feature parity with Linux ZFS.
zfs snapshot operates on datasets so if you want to use ZFS snapshots you must create at least one dataset and set a few options. Here we call our dataset main. Replace backup01 with the name of your pool and main with the name of the dataset you want to create, which you can name almost anything but its best to pick something that reflects its intended content.
$ sudo zfs create backup01/main
$ sudo zfs set compression=lz4 backup01/main
$ sudo zfs set atime=off backup01/main
$ sudo zfs set mountpoint=/backup01 backup01
You can add as many datasets as you want now or at a later date. Enabling lz4 compression and disabling updating of file access times (atime) both help improve performance. I have opted to mount this pool under /backup01 to match its pool name but you can set the pool to be mounted where you wish.
You may now check your datasets settings before exporting it:
$ zfs get all backup01/main
The ZFS pool has been created and configured along with a dataset. We can now finalise our changes to the pool by exporting it:
$ sudo zpool export backup01
Replace backup01 with the name of the pool you have just created. Exporting a ZFS pool is analagous to unmounting it, in more conventional UNIX terms. After exporting the disk containing the pool it can be safely removed from the system.
If you are using multiple ZFS pools with some on removable disks that aren't always connected to your machine then you may experience a delay during boot if you have updated your kernel when an external pool is imported that isn't always attached because it gets added to the initramfs zpool cache when a kernel update/install updates it. The fix is to clear your zpool cache and update your initramfs as described for Arch here.
ALEZ - The Arch Linux Easy ZFS installer is the easiest way to install (Arch) Linux onto a ZFS root filesystem.
The FreeBSD Handbook ZFS chapter - If you are new to ZFS, reading the FreeBSD ZFS chapter is an excellent way to learn ZFS. Most of the commands are identical under Linux.
Arch wiki ZFS page - The Arch Linux ZFS wiki page is largest Linux-specific ZFS resource online. Most of it applies to all Linux distros.
Gentoo wiki ZFS page - The gentoo wiki page is another good ZoL resource. Again, most of it applies to all Linux distros, not just gentoo.