Welcome!

Open Source Cloud Authors: Liz McMillan, Zakia Bouachraoui, Pat Romanski, Elizabeth White, Yeshim Deniz

Related Topics: Linux Containers, Open Source Cloud, Containers Expo Blog

Linux Containers: Article

ZFS on Linux

How ZFS on Linux compares to ZFS on Illumos or FreeBSD

On March 27, 2013, ZoL maintainers announced that the 0.6.1 release was ready for wide scale deployment on everything from desktops to servers. Yet, due to lack of maturity and adoption of the ZoL project, maintainers and/or advocates of ZFS aren't comfortable to run ZoL in production yet.

The reason behind the reluctance to use ZoL in production is that ZFS on Solaris took a large number of years to reach maturity and went through ups and downs of bugs related to data corruption and other issues. In the same way, ZoL will need some time to mature as product. It will take about a year to be mature as more people deploy it in production. If developers want to take advantage of ZFS, they can start rolling out less important database servers (i.e. reporting servers, 3rd slave databases) into production and experience the product for about 6 months before rolling out to all database servers. This will give users the confidence and experience to work with ZoL. Alternatively, developers may want to run ZFS on OmniOS because it's been battle tested for decades now.

How ZFS on Linux Compares to ZFS on Illumos or FreeBSD
The implementation of ZFS on Linux when compared to running ZFS on Illumos or FreeBSD is not very different from the perspective of the system administrator. The management and general usage is nearly identical. The only differences are OS specific functionality. For example, on FreeBSD if a user wants to use a zvol for swap space, he/she sets the org.freebsd:swap=on property on the zvol to turn swap on. On Linux, a developer would create a vanilla zvol and set up swap like any other partition with mkswap and swapon. Under the latest versions of all three operating systems mentioned, the Zpool version is at the same level, which is to say based on zpool v28 with additional features added by way of feature flags. They are compatible, users can create a zpool on Illumos/OmniOS, use it, export it, move the disks to a FreeBSD server, import the zpool, use it, export it, move to a Linux server, import the zpool, use it, etc. This exact scenario is something we have done at OmniTI and it worked without a hitch. One issue however is that the ACL support/usability is different on each OS so you'd the user will likely have to clean up the permissions a bit.

Caveats for Running ZFS on Arch Linux in a Production Environment
ZFS under Arch Linux is not part of the main package repository. As ZFS and its utilities are maintained by a third party, developers must rely on the third party to keep the packages up to date. One issue is that every time a new kernel is released (frequently) the ZFS kernel modules must be rebuilt as well. If the company upgrades its system (pacman -Syu) and reboots, but the ZFS modules were not recompiled well, Zpools will not initialize. This becomes especially important when developers have the rootfs under ZFS since this would leave the system unbootable and the user would be forced to recover by means of Rescue CD or, in the case of AWS, moving the EBS volumes to another instance and recovering from there. Linux does have a mechanism for automating this process other than DKMS. However, the arch zfs-modules-dkms package that provides this functionality is not kept up to date, and shouldn't be used.

Also, as briefly mentioned above, it should be noted that one cannot boot directly from ZFS on Linux, users must maintain a Linux bootloader compatible file system for /boot such as ext[234].

Currently, many of the utilities that output information about the filesystem are not ZFS aware and developers can get strange results running commands such as "df" for example, since it does not know the relationship between datasets and their parents. These will not necessarily prevent anything from running, but it is worth noting. Generally it's best to use "zfs list" rather than "df" to get accurate results.

ZFS natively uses NFSv4-style ACLs and is not compatible with Posix-style ACLs. Any applications that rely on Posix-style ACLs will have issues. Default GNU utilities like "ls" for example are not NFSv4 ACL aware.

Lastly, the ZoL project proclaims that ZFS on Linux is production ready, however it is worthwhile to note that it is still very immature at this point. ZFS itself has been around and tested for quite some time and is mature, so be careful and test before using it in a production environment.

How To Install ZFS on Arch with RootFS on ZFS
The Arch Wiki page for Installing Linux on ZFS goes into great detail on how to install Linux on ZFS. The key points are as follows:

  • The ZFS utilities and kernel modules must be built/installed prior to beginning the installation (within the CD Boot environment)
  • Even though you can have the Root FS on ZFS, the Linux bootloaders cannot load the kernel from ZFS currently so you still need a small ext2/3 partition for /boot to hold the kernel, the initramfs, and files that the bootloader requires.
  • There is no "beadm" in Linux to support multiple Boot Environment snapshots currently. One of the benefits of ZFS on Illumos/OmniOS is the ability to rollback to an earlier boot environment when applying updates.
  • When building the initramfs image, the zfs hook must come before the filesystems hook and you should not use the fsck hook at all.
  • You need to enable the ZFS service in systemd as this is not enabled by default. Under a ZFS Root system, this is very important if you like your systems to boot.
  • The "kernel" line of the bootloader needs to include a parameter telling the kernel where the root FS resides. For example, if the root FS is on a Zpool named "rpool" and its dataset is rpool/ROOT/default, then this parameter would be zfs=rpool/ROOT/default.

It is important to remember to export the zpool prior to rebooting after installation otherwise ZFS will complain that the system is different and will not import itself. This is because the new system is, in fact, a different system than the CD Media boot environment. Also, it's a very good idea to rebuild the initramfs (mkinitcpio -p linux) right away once you log into the installed system for the first time to avoid any "pool may be in use" errors due to differences in the CD Media Boot environment when the ramdisk was created initially.

Included at the end of this article are portions of a script used to build ArchLinux on ZFS. The only parts that have been removed are things that are specific to my environment. This is given as an example only to illustrate the steps that can be used, but note that it may or may not match the methods typically used for your environment.

Referencesz
ZFS on Linux Main Page: http://zfsonlinux.org

Arch Wiki - Installing Linux on ZFS: https://wiki.archlinux.org/index.php/Installing_Arch_Linux_on_ZFS

Arch Wiki - ZFS: https://wiki.archlinux.org/index.php/ZFS

Appendix A - Excerpts from Vagrant install script

The following are the commands we use when installing ZFS on ArchLinux under Vagrant. The Vagrant specific bits have been removed as they would not apply for installation on a production server. The full script can be found here: https://github.com/Loki22/scripts/blob/master/Vagrant/archzfs_vagrant_install.sh

pacman -Syy

pacman -S --noconfirm base-devel

mkdir /root/build

cd /root/build

wget https://aur.archlinux.org/packages/sp/spl-utils/spl-utils.tar.gz

wget https://aur.archlinux.org/packages/sp/spl/spl.tar.gz

wget https://aur.archlinux.org/packages/zf/zfs-utils/zfs-utils.tar.gz

wget https://aur.archlinux.org/packages/zf/zfs/zfs.tar.gz

for i in spl-utils spl zfs-utils zfs

do

cd /root/build && tar zxvf ${i}.tar.gz

cd /root/build/${i}

makepkg -s --asroot --noconfirm && pacman -U --noconfirm ./${i}*.pkg.tar.xz

done

# Install packages needed for ZFS

pacman -S --noconfirm archzfs dosfstools gptfdisk

# Clear the disk and initialize in GPT Format

sgdisk -o -g /dev/sda

# Partitioning - 3 Partitions (BIOS Boot Partition, /boot, and ZFS)

sgdisk -n 2:2048:+512M -c 2:"Linux Boot Partition" -t 2:8300 /dev/sda

sgdisk -n 3:0:0 -c 3:"ZFS Root Pool" -t 3:bf00 /dev/sda

sgdisk -n 1:34:2047 -c 1:"BIOS Boot Partition" -t 1:ef02 /dev/sda

# Create filesystem for /boot partition

mkfs.ext4 -L BOOT /dev/sda2

# Set up the ZFS Root Pool

modprobe zfs

zpool create rpool /dev/sda3

zfs set checksum=fletcher4 rpool

zfs set atime=off rpool

zfs set compression=lzjb rpool

zfs set mountpoint=none rpool

zpool export rpool

zpool import -d /dev/disk/by-id -R /mnt rpool

# Set up the initial BE (linux doesn't have beadm at this point, but not a bad idea to think ahead)

zfs create rpool/ROOT

zfs create -o mountpoint=/ rpool/ROOT/default

zpool set bootfs=rpool/ROOT/default rpool

# Set up datasets that are not part of the BE

zfs create -o mountpoint=/home -o setuid=off rpool/home

zfs create -o mountpoint=/root -o setuid=off rpool/roothome

# Create swap (example here is 2GB, use 4K block size for 64 bit systems)

zfs create -V 2G -b 4K rpool/swap

mkswap -Lswap -f /dev/rpool/swap

swapon /dev/rpool/swap

# Mount /boot

mkdir /mnt/boot

mount /dev/sda2 /mnt/boot

# Change ZFS repo to core now that we have it installed. This is so the new system will use updated modules linked to the new kernel as opposed to the somewhat more stale kernel that is used on the Install CD.

sed -i 's/demz-repo-archiso/demz-repo-core/' /etc/pacman.conf

# Bootstrap the new installation

pacstrap /mnt base base-devel archzfs sudo gnupg vim

# Generate the fstab minus the ZFS bits of which mounting is handled by ZFS

genfstab -U -p /mnt | grep boot >> /mnt/etc/fstab

# Configuration

CHROOT="arch-chroot /mnt"

# Hostname

echo "myhostname" > /mnt/etc/hostname

# Timezone and Clock

ln -s /usr/share/zoneinfo/America/New_York /mnt/etc/localtime

hwclock --systohc --utc

# Locale

sed -i 's/^#\(en_US.*\)/\1/' /mnt/etc/locale.gen

$CHROOT locale-gen

echo 'LANG="en_US.UTF-8"' > /mnt/etc/locale.conf

# Keymap

echo "KEYMAP=us" > /mnt/etc/vconsole.conf

# Mkinitcpio

sed -i 's/^\(HOOKS.*\)filesystems keyboard fsck/\1keyboard zfs filesystems/' /mnt/etc/mkinitcpio.conf

$CHROOT mkinitcpio -p linux

# Enable ZFS at boot

$CHROOT systemctl enable zfs.service

# Install GRUB

$CHROOT pacman -S --noconfirm grub-bios

modprobe dm-mod

$CHROOT grub-install --target=i386-pc --recheck --debug /dev/sda

cp /mnt/usr/share/locale/en\@quot/LC_MESSAGES/grub.mo /mnt/boot/grub/locale/en.mo

mv /mnt/boot/grub/grub.cfg /mnt/boot/grub/grub.cfg.orig

cat > /mnt/boot/grub/grub.cfg <<EOF

set timeout=2

set default=0

# (0) Arch Linux

menuentry "Arch Linux" {

set root=(hd0,2)

linux /vmlinuz-linux zfs=rpool/ROOT/default

initrd /initramfs-linux.img

}

# (1) Arch Linux (fallback)

menuentry "Arch Linux - Fallback" {

set root=(hd0,2)

linux /vmlinuz-linux zfs=rpool/ROOT/default

initrd /initramfs-linux-fallback.img

}

EOF

# SSH

$CHROOT pacman -S --noconfirm openssh

ln -s '/usr/lib/systemd/system/sshd.service' \

'/mnt/etc/systemd/system/multi-user.target.wants/sshd.service'

# Networking on installed system

# Manual linking because systemd isn't running yet

# Run 'ip link' to check the network interface and make sure it's enp0s3

ln -s '/usr/lib/systemd/system/[email protected]' \

'/mnt/etc/systemd/system/multi-user.target.wants/[email protected]'

# Clean up

# Remove downloaded packages

$CHROOT pacman -Scc --noconfirm

# Set your root password

passwd root

# Unmount filesystems, change ZFS mountpoints, and reboot

umount /mnt/boot

zfs umount -a

zpool export rpool

echo "If there were no errors, it would now be safe to reboot into the new system."

Appendix B - Recovery process if ZFS modules are not rebuilt on kernel upgrade
As mentioned above, the ZFS modules need to be rebuilt on every kernel upgrade. If this isn't done, you need to recover from a rescue environment. The recovery process (assuming booting from CD) is to build the ZFS modules/utils from the AUR (spl-utils, spl, zfs-utils, and zfs) in the temporary rescue environment, loading the ZFS module, mounting the Zpool under /mnt, mounting the /boot FS at /mnt/boot, chrooting, building the ZFS modules/utils again against the kernel in the chroot environment, rebuilding initramfs (mkinitcpio -p linux), and rebooting. Needless to say, not fun while people are screaming at you because the production server is down. This problem will be alleviated at some point when the ZFS packages are adopted into the main repositories and maintained with the rest of the release process.

More Stories By Kevin Loukinen

Kevin Loukinen is Site Reliability Engineer at OmniTI. Prior to that, he worked both as a Systems Administrator and Network Administrator for more than 12 years across several industries (financial, government and telecommunications).

Comments (0)

Share your thoughts on this story.

Add your comment
You must be signed in to add a comment. Sign-in | Register

In accordance with our Comment Policy, we encourage comments that are on topic, relevant and to-the-point. We will remove comments that include profanity, personal attacks, racial slurs, threats of violence, or other inappropriate material that violates our Terms and Conditions, and will block users who make repeated violations. We ask all readers to expect diversity of opinion and to treat one another with dignity and respect.


IoT & Smart Cities Stories
Andrew Keys is Co-Founder of ConsenSys Enterprise. He comes to ConsenSys Enterprise with capital markets, technology and entrepreneurial experience. Previously, he worked for UBS investment bank in equities analysis. Later, he was responsible for the creation and distribution of life settlement products to hedge funds and investment banks. After, he co-founded a revenue cycle management company where he learned about Bitcoin and eventually Ethereal. Andrew's role at ConsenSys Enterprise is a mul...
Whenever a new technology hits the high points of hype, everyone starts talking about it like it will solve all their business problems. Blockchain is one of those technologies. According to Gartner's latest report on the hype cycle of emerging technologies, blockchain has just passed the peak of their hype cycle curve. If you read the news articles about it, one would think it has taken over the technology world. No disruptive technology is without its challenges and potential impediments t...
Nicolas Fierro is CEO of MIMIR Blockchain Solutions. He is a programmer, technologist, and operations dev who has worked with Ethereum and blockchain since 2014. His knowledge in blockchain dates to when he performed dev ops services to the Ethereum Foundation as one the privileged few developers to work with the original core team in Switzerland.
René Bostic is the Technical VP of the IBM Cloud Unit in North America. Enjoying her career with IBM during the modern millennial technological era, she is an expert in cloud computing, DevOps and emerging cloud technologies such as Blockchain. Her strengths and core competencies include a proven record of accomplishments in consensus building at all levels to assess, plan, and implement enterprise and cloud computing solutions. René is a member of the Society of Women Engineers (SWE) and a m...
If a machine can invent, does this mean the end of the patent system as we know it? The patent system, both in the US and Europe, allows companies to protect their inventions and helps foster innovation. However, Artificial Intelligence (AI) could be set to disrupt the patent system as we know it. This talk will examine how AI may change the patent landscape in the years to come. Furthermore, ways in which companies can best protect their AI related inventions will be examined from both a US and...
In his general session at 19th Cloud Expo, Manish Dixit, VP of Product and Engineering at Dice, discussed how Dice leverages data insights and tools to help both tech professionals and recruiters better understand how skills relate to each other and which skills are in high demand using interactive visualizations and salary indicator tools to maximize earning potential. Manish Dixit is VP of Product and Engineering at Dice. As the leader of the Product, Engineering and Data Sciences team at D...
Bill Schmarzo, Tech Chair of "Big Data | Analytics" of upcoming CloudEXPO | DXWorldEXPO New York (November 12-13, 2018, New York City) today announced the outline and schedule of the track. "The track has been designed in experience/degree order," said Schmarzo. "So, that folks who attend the entire track can leave the conference with some of the skills necessary to get their work done when they get back to their offices. It actually ties back to some work that I'm doing at the University of San...
When talking IoT we often focus on the devices, the sensors, the hardware itself. The new smart appliances, the new smart or self-driving cars (which are amalgamations of many ‘things'). When we are looking at the world of IoT, we should take a step back, look at the big picture. What value are these devices providing. IoT is not about the devices, its about the data consumed and generated. The devices are tools, mechanisms, conduits. This paper discusses the considerations when dealing with the...
Bill Schmarzo, author of "Big Data: Understanding How Data Powers Big Business" and "Big Data MBA: Driving Business Strategies with Data Science," is responsible for setting the strategy and defining the Big Data service offerings and capabilities for EMC Global Services Big Data Practice. As the CTO for the Big Data Practice, he is responsible for working with organizations to help them identify where and how to start their big data journeys. He's written several white papers, is an avid blogge...
Dynatrace is an application performance management software company with products for the information technology departments and digital business owners of medium and large businesses. Building the Future of Monitoring with Artificial Intelligence. Today we can collect lots and lots of performance data. We build beautiful dashboards and even have fancy query languages to access and transform the data. Still performance data is a secret language only a couple of people understand. The more busine...