OpenZFS 2.4.4

OpenZFS is a combined filesystem and volume manager with checksummed data integrity, snapshots, compression, and native RAID. Version 2.4.4 landed alongside 2.3.9 and 2.2.11, together pulling in 967 commits. We covered it when it shipped.

Kernel support is the headline

The officially supported Linux kernel range now runs 4.18 through 7.2.

ZFS on Linux lives permanently one step behind the kernel because of licensing. The CDDL and GPL are incompatible enough that ZFS cannot be merged into the kernel tree, so it ships as an out-of-tree module that must be updated whenever the kernel changes an internal API.

The practical consequence is the one every ZFS user has hit: you update your kernel, the module fails to build, and your pool does not import on the next boot.

# Before rebooting into a new kernel
modinfo zfs | grep ^version
zpool status

On a ZFS root system, never reboot into a new kernel without confirming the module built. This is the single most important operational habit for ZFS on Linux.

The MMP recovery tool

The new zhack mmp reclaim command recovers a pool left stranded by Multi-Modifier Protection.

MMP prevents the worst thing that can happen to a pool: two hosts importing it simultaneously and corrupting it. It works by writing periodic uberblock updates so another host can tell the pool is in use. The failure mode is that a pool can look active when nothing is using it, typically after a host died uncleanly, and refuse to import.

2.4.4 also improves how OpenZFS distinguishes a failed uberblock claim from genuine activity by another host, reducing how often you reach that state.

Why people run ZFS

• Checksums on everything, with automatic repair from redundancy when corruption is detected
• Snapshots that are instant and cost only the changed blocks
• Send and receive for efficient replication to another machine
• Compression that is usually a net performance win, since less I/O beats more CPU
• RAID-Z without the write hole that afflicts traditional RAID5

The cost is memory, out-of-tree module management, and a set of concepts you have to learn properly. ARC will use available RAM by design, which alarms people who do not expect it.

Deduplication

Do not enable it. The memory cost is substantial, the benefit is narrow, and the decision is effectively permanent for a dataset. 2.4.4 fixes several dedup table bugs, which matter to the minority running it deliberately.

Installing

sudo pacman -Syu zfs-dkms                    # Arch, from AUR or archzfs
sudo dnf install zfs                         # Fedora, via the OpenZFS repo
sudo apt install zfs-dkms zfsutils-linux     # Debian and Ubuntu

DKMS rebuilds the module against your current kernel. Confirm it succeeded before rebooting.

Our ext4, Btrfs, and XFS comparison covers where the in-tree filesystems sit relative to ZFS.

Verify Your Download

Release tarballs are published with checksums and signatures on GitHub.