This repository has been archived on 2023-06-11. You can view files and clone it, but cannot push or open issues or pull requests.
blog.lazkani.io-20200902-hi.../posts/misc/a_quick_zfs_overview.rst
Elia el Lazkani 99a5aedd74
Adding the borg tag to the borgbackup post
Minor change to the name everywhere for consistency
2020-02-17 23:24:01 +01:00

173 lines
8.1 KiB
ReStructuredText
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

.. title: A quick ZFS overview on Linux
.. date: 2020-01-27
.. slug: a-quick-zfs-overview-on-linux
.. updated: 2020-01-27
.. status: published
.. tags: misc, zfs, file system
.. category: misc
.. authors: Elia el Lazkani
.. description: A overview of Linux on ZFS root
.. type: text
I have, for years, been interested in *file systems*. Specifically a *file system* to run my personal systems on.
For most people **Ext4** is good enough and that is totally fine. But, as a power user, I like to have more control, more features and more options out of my file system.
I have played with most of file sytsems on Linux, and have been using **Btrfs** for a few years now. I have worked with NAS systems running on **ZFS** and have been very impressed by it.
The only problem is that **ZFS** wasn't been well suppored on Linux at the time. **Btrfs** promissed to be the **ZFS** replacement for Linux nativetly, especially that it was backed up by a bunch of the giants like Oracle and RedHat.
My decision at that point was made, and yes that was before RedHat's support for **XFS** which is impressive on its own.
Recently though, a new project gave everyone hope. `OpenZFS <http://www.open-zfs.org/wiki/Main_Page>`_ came to life and so did `ZFS on Linux <https://zfsonlinux.org/>`_.
.. TEASER_END
Linux has had **ZFS** support for a while now but mostly to manage a **ZFS** *file system*, so I kept watching until I saw a blog post by **Ubuntu** entitled `Enhancing our ZFS support on Ubuntu 19.10 an introduction <https://ubuntu.com/blog/enhancing-our-zfs-support-on-ubuntu-19-10-an-introduction>`_.
In the blog post above, I read the following:
We want to support ZFS on root as an experimental installer option, initially for desktop,
but keeping the layout extensible for server later on. The desktop will be the first
beneficiary in Ubuntu 19.10. Note the use of the term experimental though!
My eyes widened at this point. I know that **Ubuntu** has had native **ZFS** support since 2016 but now I could install it with one click. At that point I was all in, and I went back to **Ubuntu**.
Ubuntu on root ZFS
==================
You heard me right, the **Ubuntu** installer offers an 'experimental' install on **ZFS**. I made the decision based on the well tested stability of **ZFS** in production environments and its ability to offer me the flexibility and the ability to backup and recover my data easily.
In other words, if **Ubuntu** doesn't work, **ZFS** is there and I can install whatever I like on top and if you are familiar with **ZFS** you know exactly what I mean and I have barely scratched the ice on its capabilities.
So here I was with **Ubuntu** installed on my laptop on root **ZFS**. So I had to do it.
.. code:: text
# zpool status -v
pool: bpool
state: ONLINE
status: The pool is formatted using a legacy on-disk format. The pool can
still be used, but some features are unavailable.
action: Upgrade the pool using 'zpool upgrade'. Once this is done, the
pool will no longer be accessible on software that does not support
feature flags.
scan: none requested
config:
NAME STATE READ WRITE CKSUM
bpool ONLINE 0 0 0
nvme0n1p4 ONLINE 0 0 0
errors: No known data errors
pool: rpool
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
rpool ONLINE 0 0 0
nvme0n1p5 ONLINE 0 0 0
errors: No known data errors
.. note::
I have read somewhere in a blog about **Ubuntu** that I should not run an upgrade on the boot pool.
and it's running on...
.. code:: text
# uname -s -v -i -o
Linux #28-Ubuntu SMP Wed Dec 18 05:37:46 UTC 2019 x86_64 GNU/Linux
Well that was pretty easy.
ZFS Pools
=========
Let's take a look at how the installer has configured the *pools*.
.. code:: text
# zpool list zo 23:21
NAME SIZE ALLOC FREE CKPOINT EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
bpool 1,88G 158M 1,72G - - - 8% 1.00x ONLINE -
rpool 472G 7,91G 464G - - 0% 1% 1.00x ONLINE -
So it creates a *boot* pool and a *root* pool. Maybe looking at the **datasets** would give us a better idea.
ZFS Datasets
============
Let's look at the sanitized version of the datasets.
.. code:: text
# zfs list
NAME USED AVAIL REFER MOUNTPOINT
bpool 158M 1,60G 176K /boot
bpool/BOOT 157M 1,60G 176K none
bpool/BOOT/ubuntu_xxxxxx 157M 1,60G 157M /boot
rpool 7,92G 449G 96K /
rpool/ROOT 4,53G 449G 96K none
rpool/ROOT/ubuntu_xxxxxx 4,53G 449G 3,37G /
rpool/ROOT/ubuntu_xxxxxx/srv 96K 449G 96K /srv
rpool/ROOT/ubuntu_xxxxxx/usr 208K 449G 96K /usr
rpool/ROOT/ubuntu_xxxxxx/usr/local 112K 449G 112K /usr/local
rpool/ROOT/ubuntu_xxxxxx/var 1,16G 449G 96K /var
rpool/ROOT/ubuntu_xxxxxx/var/games 96K 449G 96K /var/games
rpool/ROOT/ubuntu_xxxxxx/var/lib 1,15G 449G 1,04G /var/lib
rpool/ROOT/ubuntu_xxxxxx/var/lib/AccountServices 96K 449G 96K /var/lib/AccountServices
rpool/ROOT/ubuntu_xxxxxx/var/lib/NetworkManager 152K 449G 152K /var/lib/NetworkManager
rpool/ROOT/ubuntu_xxxxxx/var/lib/apt 75,2M 449G 75,2M /var/lib/apt
rpool/ROOT/ubuntu_xxxxxx/var/lib/dpkg 36,5M 449G 36,5M /var/lib/dpkg
rpool/ROOT/ubuntu_xxxxxx/var/log 11,0M 449G 11,0M /var/log
rpool/ROOT/ubuntu_xxxxxx/var/mail 96K 449G 96K /var/mail
rpool/ROOT/ubuntu_xxxxxx/var/snap 128K 449G 128K /var/snap
rpool/ROOT/ubuntu_xxxxxx/var/spool 112K 449G 112K /var/spool
rpool/ROOT/ubuntu_xxxxxx/var/www 96K 449G 96K /var/www
rpool/USERDATA 3,38G 449G 96K /
rpool/USERDATA/user_yyyyyy 3,37G 449G 3,37G /home/user
rpool/USERDATA/root_yyyyyy 7,52M 449G 7,52M /root
.. note::
The installer have created some random IDs that I have not figured out if they are totally random or mapped to something so I have sanitized them.
I also sanitized the user, of course. ;)
It looks like the installer created a bunch of datasets with their respective mountpoints.
ZFS Properties
==============
**ZFS** has a list of features and they are tunable in different ways, one of them is through the properties, let's have a look.
.. code:: text
# zfs get all rpool
NAME PROPERTY VALUE SOURCE
rpool type filesystem -
rpool creation vr jan 24 23:04 2020 -
rpool used 7,91G -
rpool available 449G -
rpool referenced 96K -
rpool compressratio 1.43x -
rpool mounted no -
rpool quota none default
rpool reservation none default
rpool recordsize 128K default
rpool mountpoint / local
...
This gives us an idea on properties set on the dataset specified, in this case, the *rpool* root dataset.
Conclusion
==========
I read in a blog post that the **Ubuntu** team responsible for the **ZFS** support has followed all the **ZFS** best practices in the installer.
I have no way of verifying that as I am not a **ZFS** expert but I'll be happy to take their word for it until I learn more.
What is certain for now is that I am running on **ZFS**, and I will be enjoying its features to the fullest.