Get the code: LearnZfs.txt
ZFS is a rethinking of the storage stack, combining traditional file systems as well as volume managers into one cohesive tool. ZFS has some specific terminology that sets it apart from more traditional storage systems, however it has a great set of features with a focus on usability for systems administrators.
A VDEV (Virtual Device) in ZFS is analogous to a RAID device and similarly offers different benefits in terms of redundancy and performance. In general VDEV’s offer better reliability and safety than a RAID card. It is discouraged to use a RAID setup with ZFS, as ZFS expects to directly manage the underlying disks.
| VDEV Type | Similar RAID | Notes |
|---|---|---|
| Mirror | RAID 1 | Supports n-way mirroring for redundancy. |
| raidz1 | RAID 5 | Single disk parity, offering fault tolerance of one disk failure. |
| raidz2 | RAID 6 | Two-disk parity, can tolerate two disk failures. |
| raidz3 | - | Three-disk parity, can tolerate three disk failures. |
| Disk | - | Represents a single physical disk in a VDEV. |
| File | - | File-based VDEV, not recommended for production as it adds complexity and reduces reliability. |
Data in a ZFS storage pool is striped across all VDEVs. Adding more VDEVs, Logs, or Caches can increase IOPS (Input/Output Operations Per Second), enhancing performance. It’s crucial to balance VDEVs for optimal performance and redundancy.
ZFS uses Storage Pools as an abstraction over the lower level storage provider (VDEV), allow you to separate the user visible file system from the physical layout.
ZFS datasets are analogous to traditional filesystems but with many more features. They provide many of ZFS’s advantages. Datasets support Copy on Write snapshots, quota’s, compression and de-duplication.
One directory may contain up to 2^48 files, up to 16 exabytes each. A single storage pool can contain up to 256 zettabytes (2^78) of space, and can be striped across 2^64 devices. A single host can have 2^64 storage pools. The limits are huge.
Actions:
List zpools
# Create a raidz zpool
$ zpool create zroot raidz1 gpt/zfs0 gpt/zfs1 gpt/zfs2
# List ZPools
$ zpool list
NAME SIZE ALLOC FREE EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
zroot 141G 106G 35.2G - 43% 75% 1.00x ONLINE -
# List detailed information about a specific zpool
$ zpool list -v zroot
NAME SIZE ALLOC FREE EXPANDSZ FRAG CAP DEDUP HEALTH ALTROOT
zroot 141G 106G 35.2G - 43% 75% 1.00x ONLINE -
gptid/c92a5ccf-a5bb-11e4-a77d-001b2172c655 141G 106G 35.2G - 43% 75%
Status of zpools
# Get status information about zpools
$ zpool status
pool: zroot
state: ONLINE
scan: scrub repaired 0 in 2h51m with 0 errors on Thu Oct 1 07:08:31 2015
config:
NAME STATE READ WRITE CKSUM
zroot ONLINE 0 0 0
gptid/c92a5ccf-a5bb-11e4-a77d-001b2172c655 ONLINE 0 0 0
errors: No known data errors
# Scrubbing a zpool to correct any errors
$ zpool scrub zroot
$ zpool status -v zroot
pool: zroot
state: ONLINE
scan: scrub in progress since Thu Oct 15 16:59:14 2015
39.1M scanned out of 106G at 1.45M/s, 20h47m to go
0 repaired, 0.04% done
config:
NAME STATE READ WRITE CKSUM
zroot ONLINE 0 0 0
gptid/c92a5ccf-a5bb-11e4-a77d-001b2172c655 ONLINE 0 0 0
errors: No known data errors
Properties of zpools
# Getting properties from the pool properties can be user set or system provided.
$ zpool get all zroot
NAME PROPERTY VALUE SOURCE
zroot size 141G -
zroot capacity 75% -
zroot altroot - default
zroot health ONLINE -
...
# Setting a zpool property
$ zpool set comment="Storage of mah stuff" zroot
$ zpool get comment
NAME PROPERTY VALUE SOURCE
tank comment - default
zroot comment Storage of mah stuff local
Remove zpool
$ zpool destroy test
Actions:
Create datasets
# Create dataset
$ zfs create zroot/root/data
$ mount | grep data
zroot/root/data on /data (zfs, local, nfsv4acls)
# Create child dataset
$ zfs create zroot/root/data/stuff
$ mount | grep data
zroot/root/data on /data (zfs, local, nfsv4acls)
zroot/root/data/stuff on /data/stuff (zfs, local, nfsv4acls)
# Create Volume
$ zfs create -V zroot/win_vm
$ zfs list zroot/win_vm
NAME USED AVAIL REFER MOUNTPOINT
zroot/win_vm 4.13G 17.9G 64K -
List datasets
# List all datasets
$ zfs list
NAME USED AVAIL REFER MOUNTPOINT
zroot 106G 30.8G 144K none
zroot/ROOT 18.5G 30.8G 144K none
zroot/ROOT/10.1 8K 30.8G 9.63G /
zroot/ROOT/default 18.5G 30.8G 11.2G /
zroot/backup 5.23G 30.8G 144K none
zroot/home 288K 30.8G 144K none
...
# List a specific dataset
$ zfs list zroot/home
NAME USED AVAIL REFER MOUNTPOINT
zroot/home 288K 30.8G 144K none
# List snapshots
$ zfs list -t snapshot
zroot@daily-2015-10-15 0 - 144K -
zroot/ROOT@daily-2015-10-15 0 - 144K -
zroot/ROOT/default@daily-2015-10-15 0 - 24.2G -
zroot/tmp@daily-2015-10-15 124K - 708M -
zroot/usr@daily-2015-10-15 0 - 144K -
zroot/home@daily-2015-10-15 0 - 11.9G -
zroot/var@daily-2015-10-15 704K - 1.42G -
zroot/var/log@daily-2015-10-15 192K - 828K -
zroot/var/tmp@daily-2015-10-15 0 - 152K -
Rename datasets
$ zfs rename zroot/root/home zroot/root/old_home
$ zfs rename zroot/root/new_home zroot/root/home
Delete dataset
# Datasets cannot be deleted if they have any snapshots
$ zfs destroy zroot/root/home
Get / set properties of a dataset
# Get all properties
$ zfs get all zroot/usr/home
NAME PROPERTY VALUE SOURCE
zroot/home type filesystem -
zroot/home creation Mon Oct 20 14:44 2014 -
zroot/home used 11.9G -
zroot/home available 94.1G -
zroot/home referenced 11.9G -
zroot/home mounted yes -
...
# Get property from dataset
$ zfs get compression zroot/usr/home
NAME PROPERTY VALUE SOURCE
zroot/home compression off default
# Set property on dataset
$ zfs set compression=lz4 zroot/lamb
# Get a set of properties from all datasets
$ zfs list -o name,quota,reservation
NAME QUOTA RESERV
zroot none none
zroot/ROOT none none
zroot/ROOT/default none none
zroot/tmp none none
zroot/usr none none
zroot/home none none
zroot/var none none
...
The ZFS Intent Log (ZIL) is a write log designed to speed up synchronous writes. This is typically a faster drive or drive partition than the larger storage pools.
# Add a log pool
$ zpool add mypool/lamb log /dev/sdX
# Check the configuration
$ zpool status mypool/lamb
The Level 2 Adaptive Replacement Cache (L2ARC) extends the primary ARC (in-RAM cache) and is used for read caching. This is typically a faster drive or drive partition than the larger storage pools.
# Add a cache pool
$ zpool add mypool/lamb cache /dev/sdY
# Check the configuration
$ zpool status mypool/lamb
Data compression reduces the amount of space data occupies on disk in exchange for some extra CPU usage. When enabled, it can enhance performance by reducing the amount of disk I/O. It especially beneficial on systems with more CPU resources than disk bandwidth.
# Get compression options
$ zfs get -help
...
compression NO YES on | off | lzjb | gzip | gzip-[1-9] | zle | lz4 | zstd | zstd-[1-19] | zstd-fast | zstd-fast-[1-10,20,30,40,50,60,70,80,90,100,500,1000]
...
# Set compression
$ zfs set compression=on mypool/lamb
# Check the configuration
$ zpool get compression mypool/lamb
Encryption allows data to be encrypted on the device at the cost of extra CPU cycles. This property can only be set when a dataset is being created.
# Enable encryption on the pool
$ zpool set feature@encryption=enabled black_hole
# Create an encrypted dataset with a prompt
$ zfs create -o encryption=on -o keyformat=passphrase black_hole/enc
# Check the configuration
$ zfs get encryption black_hole/enc
It should be noted that there are parts of the system where the data is not encrypted. See the table below for a breakdown.
| Component | Encrypted | Notes |
|---|---|---|
| Main Data Storage | Yes | Data in datasets/volumes is encrypted. |
| ZFS Intent Log (ZIL) | Yes | Synchronous write requests are encrypted. |
| L2ARC (Cache) | Yes | Cached data is stored in an encrypted form. |
| RAM (ARC) | No | Data in the primary ARC, in RAM, is not encrypted. |
| Swap Area | Conditional | Encrypted if the ZFS swap dataset is encrypted. |
| ZFS Metadata | Yes | Metadata is encrypted for encrypted datasets. |
| Snapshot Data | Yes | Snapshots of encrypted datasets are also encrypted. |
| ZFS Send/Receive | Conditional | Encrypted during send/receive if datasets are encrypted and -w flag is used. |
ZFS snapshots are one of the things about zfs that are a really big deal
Actions:
Create snapshots
# Create a snapshot of a single dataset
zfs snapshot zroot/home/sarlalian@now
# Create a snapshot of a dataset and its children
$ zfs snapshot -r zroot/home@now
$ zfs list -t snapshot
NAME USED AVAIL REFER MOUNTPOINT
zroot/home@now 0 - 26K -
zroot/home/sarlalian@now 0 - 259M -
zroot/home/alice@now 0 - 156M -
zroot/home/bob@now 0 - 156M -
...
Destroy snapshots
# How to destroy a snapshot
$ zfs destroy zroot/home/sarlalian@now
# Delete a snapshot on a parent dataset and its children
$ zfs destroy -r zroot/home/sarlalian@now
Renaming Snapshots
# Rename a snapshot
$ zfs rename zroot/home/sarlalian@now zroot/home/sarlalian@today
$ zfs rename zroot/home/sarlalian@now today
$ zfs rename -r zroot/home@now @yesterday
Accessing snapshots
# CD into a snapshot directory
$ cd /home/.zfs/snapshot/
Sending and Receiving
# Backup a snapshot to a file
$ zfs send zroot/home/sarlalian@now | gzip > backup_file.gz
# Send a snapshot to another dataset
$ zfs send zroot/home/sarlalian@now | zfs recv backups/home/sarlalian
# Send a snapshot to a remote host
$ zfs send zroot/home/sarlalian@now | ssh root@backup_server 'zfs recv zroot/home/sarlalian'
# Send full dataset with snapshots to new host
$ zfs send -v -R zroot/home@now | ssh root@backup_server 'zfs recv zroot/home'
Cloning Snapshots
# Clone a snapshot
$ zfs clone zroot/home/sarlalian@now zroot/home/sarlalian_new
# Promoting the clone so it is no longer dependent on the snapshot
$ zfs promote zroot/home/sarlalian_new
This following a script utilizing FreeBSD, jails and ZFS to automate provisioning a clean copy of a MySQL staging database from a live replication slave.
#!/bin/sh
echo "==== Stopping the staging database server ===="
jail -r staging
echo "==== Cleaning up existing staging server and snapshot ===="
zfs destroy -r zroot/jails/staging
zfs destroy zroot/jails/slave@staging
echo "==== Quiescing the slave database ===="
echo "FLUSH TABLES WITH READ LOCK;" | /usr/local/bin/mysql -u root -pmyrootpassword -h slave
echo "==== Snapshotting the slave db filesystem as zroot/jails/slave@staging ===="
zfs snapshot zroot/jails/slave@staging
echo "==== Starting the slave database server ===="
jail -c slave
echo "==== Cloning the slave snapshot to the staging server ===="
zfs clone zroot/jails/slave@staging zroot/jails/staging
echo "==== Installing the staging mysql config ===="
mv /jails/staging/usr/local/etc/my.cnf /jails/staging/usr/local/etc/my.cnf.slave
cp /jails/staging/usr/local/etc/my.cnf.staging /jails/staging/usr/local/etc/my.cnf
echo "==== Setting up the staging rc.conf file ===="
mv /jails/staging/etc/rc.conf.local /jails/staging/etc/rc.conf.slave
mv /jails/staging/etc/rc.conf.staging /jails/staging/etc/rc.conf.local
echo "==== Starting the staging db server ===="
jail -c staging
echo "==== Makes the staging database not pull from the master ===="
echo "STOP SLAVE;" | /usr/local/bin/mysql -u root -pmyrootpassword -h staging
echo "RESET SLAVE;" | /usr/local/bin/mysql -u root -pmyrootpassword -h staging
Got a suggestion? A correction, perhaps? Open an Issue on the GitHub Repo, or make a pull request yourself!
Originally contributed by sarlalian, and updated by 11 contributors.