bionindo.blogg.se - Openzfs vs zfs on linux

Openzfs vs zfs on linux drivers#
Openzfs vs zfs on linux driver#
Openzfs vs zfs on linux software#
Openzfs vs zfs on linux code#
Openzfs vs zfs on linux license#

There are several Ars Technica links in this blog post. It is still a WIP as of August 2021 despite a breathless Ars Technica article about it in June. Growing a RAIDZ vdev by adding disks is at least coming soon. (Just for comparison, mdadm lets you grow a RAID volume by adding disks since 2006 and shrink by removing disks since 2009.) This could badly affect your storage efficiency. You can’t shrink a RAIDZ vdev by removing disks and you can’t grow a RAIDZ vdev in by adding disks.Īll you can do in ZFS is expand your pool to create a whole second RAIDZ vdev and stripe your pool across it, creating a RAID60 - you can’t just have one big RAID6. You can’t stack vdevs in any other configuration.įor argument’s sake, let’s assume most small installations would have a pool with only a single RAID-Z2 vdev. This RAID-X0 (stripe of mirrors) structure is rigid, you can’t do 0X (mirror of stripes) instead at all. It can be a passthrough of a single real block device, or a RAID of multiple underlying block devices.

A “vdev” is the 2nd-level block device.

It is a union span (call it RAID-0, stripe, or JBOD if you like) of all the vdevs in the pool.

A “pool” is the top-level block device.

It might be the main filesystem or perhaps a snapshot.

A “dataset” is a filesystem you can mount.

To first clarify ZFS’s custom terminology (not considered a really bad point because even LVM2 is also guilty of using custom terminology here): ZoL did workaround the Linux symbol issue above by disabling all use of SIMD for encryption, reducing the performance versus an in-tree filesystem. You could consider trying the fuse ZFS instead of the in-kernel one at least, as a userspace program it is definitely not a combined work. Red Hat will not touch this with a bargepole. This redistribution of a combined CDDL/GPLv2 work is probably illegal. Ubuntu ships ZFS as part of the kernel, not even as a separate loadable module.

Openzfs vs zfs on linux code#

They ship only ZoL’s source code with a nice script that compiles everything on your own machine ( zfs-dkms), so it is technically never redistributed in this form which satisfies both licenses.

Openzfs vs zfs on linux license#

It’s all open source, but your right to redistribute the work to others requires your compliance with both the CDDL and GPL license which can’t be satisifed simultaneously. When compiled together, loaded and runnning, the resulting kernel is a combined work of both GPL and CDDL code.

Openzfs vs zfs on linux software#

Therefore just like how all proprietary software eventually finds a GPL implementation and then a BSD/MIT one, ZFS will eventually be superceded by a mainline solution, so don’t get too used to it.Īs an out-of-tree GPL-incompatible module, it is regularly broken by upstream changes on Linux where ZoL was discovered to be abusing GPL symbols, causing long periods of unavailability until a workaround can be found.

This is not possible because ZFS is under the CDDL license and Oracle are unlikely to relicense it, if they are even legally able to.

Openzfs vs zfs on linux drivers#

Linux drivers are best maintained when they’re in the Linux kernel git repository along with all the other filesystem drivers. Bad things about ZFS Out-of-tree and will never be mainlined Scrubbing is the process of periodically, preemtively reading every sector to check for UREs and repair them early. The RAID layer can do this automatically by relying on its redundant copy. What is “Scrubbing”? If a disk has an unrecoverable read error (URE) when reading a sector, it’s possible to repair the sector by rewriting its contents the physical disk detects the rewrite over an unreadable sector and performs remapping in firmware.

Openzfs vs zfs on linux driver#

It was renamed to OpenZFS since ZoL got FreeBSD support, and FreeBSD’s own in-tree ZFS driver was deprecated in favor of just periodically syncing ZoL from out-of-tree. In this post I will refer to the ZFS on Linux project as ZoL. It is not the only way to achieve those features on Linux, and there are better alternatives. It has many good features, which are probably why it is used: snapshots (with send/receive suppport), checksumming, RAID of some kind (with scrubbing support), deduplication, compression, and encryption.īut ZFS also has a lot of downsides. ZFS is a hybrid filesystem and volume manager system that is quite popular recently but has some important and unexpected problems.