zfs: add some emphasis on *not* upgrading features

[pve-docs.git] / local-zfs.adoc

[[chapter_zfs]]
ZFS on Linux
------------
ifdef::wiki[]
:pve-toplevel:
endif::wiki[]

ZFS is a combined file system and logical volume manager designed by
Sun Microsystems. Starting with {pve} 3.4, the native Linux
kernel port of the ZFS file system is introduced as optional
file system and also as an additional selection for the root
file system. There is no need for manually compile ZFS modules - all
packages are included.

By using ZFS, its possible to achieve maximum enterprise features with
low budget hardware, but also high performance systems by leveraging
SSD caching or even SSD only setups. ZFS can replace cost intense
hardware raid cards by moderate CPU and memory load combined with easy
management.

.General ZFS advantages

* Easy configuration and management with {pve} GUI and CLI.

* Reliable

* Protection against data corruption

* Data compression on file system level

* Snapshots

* Copy-on-write clone

* Various raid levels: RAID0, RAID1, RAID10, RAIDZ-1, RAIDZ-2 and RAIDZ-3

* Can use SSD for cache

* Self healing

* Continuous integrity checking

* Designed for high storage capacities

* Asynchronous replication over network

* Open Source

* Encryption

* ...


Hardware
~~~~~~~~

ZFS depends heavily on memory, so you need at least 8GB to start. In
practice, use as much you can get for your hardware/budget. To prevent
data corruption, we recommend the use of high quality ECC RAM.

If you use a dedicated cache and/or log disk, you should use an
enterprise class SSD (e.g. Intel SSD DC S3700 Series). This can
increase the overall performance significantly.

IMPORTANT: Do not use ZFS on top of hardware controller which has its
own cache management. ZFS needs to directly communicate with disks. An
HBA adapter is the way to go, or something like LSI controller flashed
in ``IT'' mode.

If you are experimenting with an installation of {pve} inside a VM
(Nested Virtualization), don't use `virtio` for disks of that VM,
since they are not supported by ZFS. Use IDE or SCSI instead (works
also with `virtio` SCSI controller type).


Installation as Root File System
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

When you install using the {pve} installer, you can choose ZFS for the
root file system. You need to select the RAID type at installation
time:

[horizontal]
RAID0:: Also called ``striping''. The capacity of such volume is the sum
of the capacities of all disks. But RAID0 does not add any redundancy,
so the failure of a single drive makes the volume unusable.

RAID1:: Also called ``mirroring''. Data is written identically to all
disks. This mode requires at least 2 disks with the same size. The
resulting capacity is that of a single disk.

RAID10:: A combination of RAID0 and RAID1. Requires at least 4 disks.

RAIDZ-1:: A variation on RAID-5, single parity. Requires at least 3 disks.

RAIDZ-2:: A variation on RAID-5, double parity. Requires at least 4 disks.

RAIDZ-3:: A variation on RAID-5, triple parity. Requires at least 5 disks.

The installer automatically partitions the disks, creates a ZFS pool
called `rpool`, and installs the root file system on the ZFS subvolume
`rpool/ROOT/pve-1`.

Another subvolume called `rpool/data` is created to store VM
images. In order to use that with the {pve} tools, the installer
creates the following configuration entry in `/etc/pve/storage.cfg`:

----
zfspool: local-zfs
	pool rpool/data
	sparse
	content images,rootdir
----

After installation, you can view your ZFS pool status using the
`zpool` command:

----
# zpool status
  pool: rpool
 state: ONLINE
  scan: none requested
config:

	NAME        STATE     READ WRITE CKSUM
	rpool       ONLINE       0     0     0
	  mirror-0  ONLINE       0     0     0
	    sda2    ONLINE       0     0     0
	    sdb2    ONLINE       0     0     0
	  mirror-1  ONLINE       0     0     0
	    sdc     ONLINE       0     0     0
	    sdd     ONLINE       0     0     0

errors: No known data errors
----

The `zfs` command is used configure and manage your ZFS file
systems. The following command lists all file systems after
installation:

----
# zfs list
NAME               USED  AVAIL  REFER  MOUNTPOINT
rpool             4.94G  7.68T    96K  /rpool
rpool/ROOT         702M  7.68T    96K  /rpool/ROOT
rpool/ROOT/pve-1   702M  7.68T   702M  /
rpool/data          96K  7.68T    96K  /rpool/data
rpool/swap        4.25G  7.69T    64K  -
----


[[sysadmin_zfs_raid_considerations]]
ZFS RAID Level Considerations
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

There are a few factors to take into consideration when choosing the layout of
a ZFS pool. The basic building block of a ZFS pool is the virtual device, or
`vdev`. All vdevs in a pool are used equally and the data is striped among them
(RAID0). Check the `zpool(8)` manpage for more details on vdevs.

[[sysadmin_zfs_raid_performance]]
Performance
^^^^^^^^^^^

Each `vdev` type has different performance behaviors. The two
parameters of interest are the IOPS (Input/Output Operations per Second) and
the bandwidth with which data can be written or read.

A 'mirror' vdev (RAID1) will approximately behave like a single disk in regards
to both parameters when writing data. When reading data if will behave like the
number of disks in the mirror.

A common situation is to have 4 disks. When setting it up as 2 mirror vdevs
(RAID10) the pool will have the write characteristics as two single disks in
regard of IOPS and bandwidth. For read operations it will resemble 4 single
disks.

A 'RAIDZ' of any redundancy level will approximately behave like a single disk
in regard of IOPS with a lot of bandwidth. How much bandwidth depends on the
size of the RAIDZ vdev and the redundancy level.

For running VMs, IOPS is the more important metric in most situations.


[[sysadmin_zfs_raid_size_space_usage_redundancy]]
Size, Space usage and Redundancy
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

While a pool made of 'mirror' vdevs will have the best performance
characteristics, the usable space will be 50% of the disks available. Less if a
mirror vdev consists of more than 2 disks, for example in a 3-way mirror. At
least one healthy disk per mirror is needed for the pool to stay functional.

The usable space of a 'RAIDZ' type vdev of N disks is roughly N-P, with P being
the RAIDZ-level. The RAIDZ-level indicates how many arbitrary disks can fail
without losing data. A special case is a 4 disk pool with RAIDZ2. In this
situation it is usually better to use 2 mirror vdevs for the better performance
as the usable space will be the same.

Another important factor when using any RAIDZ level is how ZVOL datasets, which
are used for VM disks, behave. For each data block the pool needs parity data
which is at least the size of the minimum block size defined by the `ashift`
value of the pool. With an ashift of 12 the block size of the pool is 4k.  The
default block size for a ZVOL is 8k. Therefore, in a RAIDZ2 each 8k block
written will cause two additional 4k parity blocks to be written,
8k + 4k + 4k = 16k.  This is of course a simplified approach and the real
situation will be slightly different with metadata, compression and such not
being accounted for in this example.

This behavior can be observed when checking the following properties of the
ZVOL:

 * `volsize`
 * `refreservation` (if the pool is not thin provisioned)
 * `used` (if the pool is thin provisioned and without snapshots present)

----
# zfs get volsize,refreservation,used <pool>/vm-<vmid>-disk-X
----

`volsize` is the size of the disk as it is presented to the VM, while
`refreservation` shows the reserved space on the pool which includes the
expected space needed for the parity data. If the pool is thin provisioned, the
`refreservation` will be set to 0. Another way to observe the behavior is to
compare the used disk space within the VM and the `used` property. Be aware
that snapshots will skew the value.

There are a few options to counter the increased use of space:

* Increase the `volblocksize` to improve the data to parity ratio
* Use 'mirror' vdevs instead of 'RAIDZ'
* Use `ashift=9` (block size of 512 bytes)

The `volblocksize` property can only be set when creating a ZVOL. The default
value can be changed in the storage configuration. When doing this, the guest
needs to be tuned accordingly and depending on the use case, the problem of
write amplification if just moved from the ZFS layer up to the guest.

Using `ashift=9` when creating the pool can lead to bad
performance, depending on the disks underneath, and cannot be changed later on.

Mirror vdevs (RAID1, RAID10) have favorable behavior for VM workloads. Use
them, unless your environment has specific needs and characteristics where
RAIDZ performance characteristics are acceptable.


Bootloader
~~~~~~~~~~

Depending on whether the system is booted in EFI or legacy BIOS mode the
{pve} installer sets up either `grub` or `systemd-boot` as main bootloader.
See the chapter on xref:sysboot[{pve} host bootladers] for details.


ZFS Administration
~~~~~~~~~~~~~~~~~~

This section gives you some usage examples for common tasks. ZFS
itself is really powerful and provides many options. The main commands
to manage ZFS are `zfs` and `zpool`. Both commands come with great
manual pages, which can be read with:

----
# man zpool
# man zfs
-----

[[sysadmin_zfs_create_new_zpool]]
Create a new zpool
^^^^^^^^^^^^^^^^^^

To create a new pool, at least one disk is needed. The `ashift` should
have the same sector-size (2 power of `ashift`) or larger as the
underlying disk.

----
# zpool create -f -o ashift=12 <pool> <device>
----

To activate compression (see section <<zfs_compression,Compression in ZFS>>):

----
# zfs set compression=lz4 <pool>
----

[[sysadmin_zfs_create_new_zpool_raid0]]
Create a new pool with RAID-0
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Minimum 1 disk

----
# zpool create -f -o ashift=12 <pool> <device1> <device2>
----

[[sysadmin_zfs_create_new_zpool_raid1]]
Create a new pool with RAID-1
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Minimum 2 disks

----
# zpool create -f -o ashift=12 <pool> mirror <device1> <device2>
----

[[sysadmin_zfs_create_new_zpool_raid10]]
Create a new pool with RAID-10
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Minimum 4 disks

----
# zpool create -f -o ashift=12 <pool> mirror <device1> <device2> mirror <device3> <device4>
----

[[sysadmin_zfs_create_new_zpool_raidz1]]
Create a new pool with RAIDZ-1
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Minimum 3 disks

----
# zpool create -f -o ashift=12 <pool> raidz1 <device1> <device2> <device3>
----

Create a new pool with RAIDZ-2
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Minimum 4 disks

----
# zpool create -f -o ashift=12 <pool> raidz2 <device1> <device2> <device3> <device4>
----

[[sysadmin_zfs_create_new_zpool_with_cache]]
Create a new pool with cache (L2ARC)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

It is possible to use a dedicated cache drive partition to increase
the performance (use SSD).

As `<device>` it is possible to use more devices, like it's shown in
"Create a new pool with RAID*".

----
# zpool create -f -o ashift=12 <pool> <device> cache <cache_device>
----

[[sysadmin_zfs_create_new_zpool_with_log]]
Create a new pool with log (ZIL)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

It is possible to use a dedicated cache drive partition to increase
the performance(SSD).

As `<device>` it is possible to use more devices, like it's shown in
"Create a new pool with RAID*".

----
# zpool create -f -o ashift=12 <pool> <device> log <log_device>
----

[[sysadmin_zfs_add_cache_and_log_dev]]
Add cache and log to an existing pool
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

If you have a pool without cache and log. First partition the SSD in
2 partition with `parted` or `gdisk`

IMPORTANT: Always use GPT partition tables.

The maximum size of a log device should be about half the size of
physical memory, so this is usually quite small. The rest of the SSD
can be used as cache.

----
# zpool add -f <pool> log <device-part1> cache <device-part2>
----

[[sysadmin_zfs_change_failed_dev]]
Changing a failed device
^^^^^^^^^^^^^^^^^^^^^^^^

----
# zpool replace -f <pool> <old device> <new device>
----

.Changing a failed bootable device

Depending on how {pve} was installed it is either using `grub` or `systemd-boot`
as bootloader (see xref:sysboot[Host Bootloader]).

The first steps of copying the partition table, reissuing GUIDs and replacing
the ZFS partition are the same. To make the system bootable from the new disk,
different steps are needed which depend on the bootloader in use.

----
# sgdisk <healthy bootable device> -R <new device>
# sgdisk -G <new device>
# zpool replace -f <pool> <old zfs partition> <new zfs partition>
----

NOTE: Use the `zpool status -v` command to monitor how far the resilvering
process of the new disk has progressed.

.With `systemd-boot`:

----
# pve-efiboot-tool format <new disk's ESP>
# pve-efiboot-tool init <new disk's ESP>
----

NOTE: `ESP` stands for EFI System Partition, which is setup as partition #2 on
bootable disks setup by the {pve} installer since version 5.4. For details, see
xref:sysboot_systemd_boot_setup[Setting up a new partition for use as synced ESP].

.With `grub`:

----
# grub-install <new disk>
----

Activate E-Mail Notification
~~~~~~~~~~~~~~~~~~~~~~~~~~~~

ZFS comes with an event daemon, which monitors events generated by the
ZFS kernel module. The daemon can also send emails on ZFS events like
pool errors. Newer ZFS packages ship the daemon in a separate package,
and you can install it using `apt-get`:

----
# apt-get install zfs-zed
----

To activate the daemon it is necessary to edit `/etc/zfs/zed.d/zed.rc` with your
favourite editor, and uncomment the `ZED_EMAIL_ADDR` setting:

--------
ZED_EMAIL_ADDR="root"
--------

Please note {pve} forwards mails to `root` to the email address
configured for the root user.

IMPORTANT: The only setting that is required is `ZED_EMAIL_ADDR`. All
other settings are optional.


[[sysadmin_zfs_limit_memory_usage]]
Limit ZFS Memory Usage
~~~~~~~~~~~~~~~~~~~~~~

ZFS uses '50 %' of the host memory for the **A**daptive **R**eplacement
**C**ache (ARC) by default. Allocating enough memory for the ARC is crucial for
IO performance, so reduce it with caution. As a general rule of thumb, allocate
at least +2 GiB Base + 1 GiB/TiB-Storage+. For example, if you have a pool with
+8 TiB+ of available storage space then you should use +10 GiB+ of memory for
the ARC.

You can change the ARC usage limit for the current boot (a reboot resets this
change again) by writing to the +zfs_arc_max+ module parameter directly:

----
 echo "$[10 * 1024*1024*1024]" >/sys/module/zfs/parameters/zfs_arc_max
----

To *permanently change* the ARC limits, add the following line to
`/etc/modprobe.d/zfs.conf`:

--------
options zfs zfs_arc_max=8589934592
--------

This example setting limits the usage to 8 GiB ('8 * 2^30^').

[IMPORTANT]
====
If your root file system is ZFS, you must update your initramfs every
time this value changes:

----
# update-initramfs -u
----

You *must reboot* to activate these changes.
====


[[zfs_swap]]
SWAP on ZFS
~~~~~~~~~~~

Swap-space created on a zvol may generate some troubles, like blocking the
server or generating a high IO load, often seen when starting a Backup
to an external Storage.

We strongly recommend to use enough memory, so that you normally do not
run into low memory situations. Should you need or want to add swap, it is
preferred to create a partition on a physical disk and use it as swapdevice.
You can leave some space free for this purpose in the advanced options of the
installer. Additionally, you can lower the
``swappiness'' value. A good value for servers is 10:

----
# sysctl -w vm.swappiness=10
----

To make the swappiness persistent, open `/etc/sysctl.conf` with
an editor of your choice and add the following line:

--------
vm.swappiness = 10
--------

.Linux kernel `swappiness` parameter values
[width="100%",cols="<m,2d",options="header"]
|===========================================================
| Value               | Strategy
| vm.swappiness = 0   | The kernel will swap only to avoid
an 'out of memory' condition
| vm.swappiness = 1   | Minimum amount of swapping without
disabling it entirely.
| vm.swappiness = 10  | This value is sometimes recommended to
improve performance when sufficient memory exists in a system.
| vm.swappiness = 60  | The default value.
| vm.swappiness = 100 | The kernel will swap aggressively.
|===========================================================

[[zfs_encryption]]
Encrypted ZFS Datasets
~~~~~~~~~~~~~~~~~~~~~~

ZFS on Linux version 0.8.0 introduced support for native encryption of
datasets. After an upgrade from previous ZFS on Linux versions, the encryption
feature can be enabled per pool:

----
# zpool get feature@encryption tank
NAME  PROPERTY            VALUE            SOURCE
tank  feature@encryption  disabled         local

# zpool set feature@encryption=enabled

# zpool get feature@encryption tank
NAME  PROPERTY            VALUE            SOURCE
tank  feature@encryption  enabled         local
----

WARNING: There is currently no support for booting from pools with encrypted
datasets using Grub, and only limited support for automatically unlocking
encrypted datasets on boot. Older versions of ZFS without encryption support
will not be able to decrypt stored data.

NOTE: It is recommended to either unlock storage datasets manually after
booting, or to write a custom unit to pass the key material needed for
unlocking on boot to `zfs load-key`.

WARNING: Establish and test a backup procedure before enabling encryption of
production data. If the associated key material/passphrase/keyfile has been
lost, accessing the encrypted data is no longer possible.

Encryption needs to be setup when creating datasets/zvols, and is inherited by
default to child datasets. For example, to create an encrypted dataset
`tank/encrypted_data` and configure it as storage in {pve}, run the following
commands:

----
# zfs create -o encryption=on -o keyformat=passphrase tank/encrypted_data
Enter passphrase:
Re-enter passphrase:

# pvesm add zfspool encrypted_zfs -pool tank/encrypted_data
----

All guest volumes/disks create on this storage will be encrypted with the
shared key material of the parent dataset.

To actually use the storage, the associated key material needs to be loaded
and the dataset needs to be mounted. This can be done in one step with:

----
# zfs mount -l tank/encrypted_data
Enter passphrase for 'tank/encrypted_data':
----

It is also possible to use a (random) keyfile instead of prompting for a
passphrase by setting the `keylocation` and `keyformat` properties, either at
creation time or with `zfs change-key` on existing datasets:

----
# dd if=/dev/urandom of=/path/to/keyfile bs=32 count=1

# zfs change-key -o keyformat=raw -o keylocation=file:///path/to/keyfile tank/encrypted_data
----

WARNING: When using a keyfile, special care needs to be taken to secure the
keyfile against unauthorized access or accidental loss. Without the keyfile, it
is not possible to access the plaintext data!

A guest volume created underneath an encrypted dataset will have its
`encryptionroot` property set accordingly. The key material only needs to be
loaded once per encryptionroot to be available to all encrypted datasets
underneath it.

See the `encryptionroot`, `encryption`, `keylocation`, `keyformat` and
`keystatus` properties, the `zfs load-key`, `zfs unload-key` and `zfs
change-key` commands and the `Encryption` section from `man zfs` for more
details and advanced usage.


[[zfs_compression]]
Compression in ZFS
~~~~~~~~~~~~~~~~~~

When compression is enabled on a dataset, ZFS tries to compress all *new*
blocks before writing them and decompresses them on reading. Already
existing data will not be compressed retroactively.

You can enable compression with:

----
# zfs set compression=<algorithm> <dataset>
----

We recommend using the `lz4` algorithm, because it adds very little CPU
overhead. Other algorithms like `lzjb` and `gzip-N`, where `N` is an
integer from `1` (fastest) to `9` (best compression ratio), are also
available. Depending on the algorithm and how compressible the data is,
having compression enabled can even increase I/O performance.

You can disable compression at any time with:

----
# zfs set compression=off <dataset>
----

Again, only new blocks will be affected by this change.


[[sysadmin_zfs_special_device]]
ZFS Special Device
~~~~~~~~~~~~~~~~~~

Since version 0.8.0 ZFS supports `special` devices. A `special` device in a
pool is used to store metadata, deduplication tables, and optionally small
file blocks.

A `special` device can improve the speed of a pool consisting of slow spinning
hard disks with a lot of metadata changes. For example workloads that involve
creating, updating or deleting a large number of files will benefit from the
presence of a `special` device. ZFS datasets can also be configured to store
whole small files on the `special` device which can further improve the
performance. Use fast SSDs for the `special` device.

IMPORTANT: The redundancy of the `special` device should match the one of the
pool, since the `special` device is a point of failure for the whole pool.

WARNING: Adding a `special` device to a pool cannot be undone!

.Create a pool with `special` device and RAID-1:

----
# zpool create -f -o ashift=12 <pool> mirror <device1> <device2> special mirror <device3> <device4>
----

.Add a `special` device to an existing pool with RAID-1:

----
# zpool add <pool> special mirror <device1> <device2>
----

ZFS datasets expose the `special_small_blocks=<size>` property. `size` can be
`0` to disable storing small file blocks on the `special` device or a power of
two in the range between `512B` to `128K`. After setting the property new file
blocks smaller than `size` will be allocated on the `special` device.

IMPORTANT: If the value for `special_small_blocks` is greater than or equal to
the `recordsize` (default `128K`) of the dataset, *all* data will be written to
the `special` device, so be careful!

Setting the `special_small_blocks` property on a pool will change the default
value of that property for all child ZFS datasets (for example all containers
in the pool will opt in for small file blocks).

.Opt in for all file smaller than 4K-blocks pool-wide:

----
# zfs set special_small_blocks=4K <pool>
----

.Opt in for small file blocks for a single dataset:

----
# zfs set special_small_blocks=4K <pool>/<filesystem>
----

.Opt out from small file blocks for a single dataset:

----
# zfs set special_small_blocks=0 <pool>/<filesystem>
----

[[sysadmin_zfs_features]]
ZFS Pool Features
~~~~~~~~~~~~~~~~~

Changes to the on-disk format in ZFS are only made between major version changes
and are specified through *features*. All features, as well as the general
mechanism are well documented in the `zpool-features(5)` manpage.

Since enabling new features can render a pool not importable by an older version
of ZFS, this needs to be done actively by the administrator, by running
`zpool upgrade` on the pool (see the `zpool-upgrade(8)` manpage).

Unless you need to use one of the new features, there is no upside to enabling
them.

In fact, there are some downsides to enabling new features:

* A system with root on ZFS, that still boots using `grub` will become
  unbootable if a new feature is active on the rpool, due to the incompatible
  implementation of ZFS in grub.
* The system will not be able to import any upgraded pool when booted with an
  older kernel, which still ships with the old ZFS modules.
* Booting an older {pve} ISO to repair a non-booting system will likewise not
  work.

IMPORTANT: Do *not* upgrade your rpool if your system is still booted with
`grub`, as this will render your system unbootable. This includes systems
installed before {pve} 5.4, and systems booting with legacy BIOS boot (see
xref:sysboot_determine_bootloader_used[how to determine the bootloader]).

.Enable new features for a ZFS pool:
----
# zpool upgrade <pool>
----