Difference between revisions of "Storage: ZFS"

From Proxmox VE
Jump to navigation Jump to search
(move user contrib docs to ZFS: Tips and Tricks)
Line 1: Line 1:
== Introduction ==
ZFS is a combined file system and logical volume manager designed by Sun Microsystems. Starting with Proxmox VE 3.4, the native Linux kernel port of the ZFS filesystem is introduced as optional file-system and also as an additional selection for the root file-system. There is no need for manually compiling ZFS, all packages are included (for both kernel branches, 2.6.32 and 3.10).
By using ZFS, its possible to achieve maximal 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.
In the first release, there are two ways to use ZFS on Proxmox VE:
*as an local directory, supports all storage content types (instead of ext3 or ext4)
*as zvol block-storage, currently supporting kvm images in raw format (new ZFS storage plugin)
**The advantage of zvol is the snapshot capability on fs-level (fast)
This articles describes how to use ZFS on Proxmox VE.
=== General ZFS advantages ===
*Easy configuration and management with Proxmox VE GUI and CLI.
*Protection against data corruption
*Data compression on file-system level
*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
*Protection against data corruption
*Asynchrony replication over network
*Open Source
== Hardware ==
ZFS depends heavily on memory, so you need at least 4GB to start. In practice, use as much you can get for your hardware/budget. To prevent data corruption, the use of high quality ECC RAM is very recommended.
If you use a dedicated cache and/or log disk, you should use a enterprise class SSD (e.g. Intel SSD DC S3700 Series). This can increase the overall performance quite significantly.
If you are experimenting with an installation of Proxmox inside a VM ([[Nested_Virtualization]]), don't use Virtio for disks of that VM, since are not supported by ZFS, use IDE or SCSI (it's ok also with Virtio type contorller) instead.
<b>IMPORTANT:</b> Do not use zfs on top of hardware controller which has it's 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).
== Installation as root file-system ==
When you install with Proxmox-VE installer grater than 3.4. you can choose what FS you prefer.
It is not possible to use ZFS as rpool(root partition) with UEFI boot.
== Administration ==
=== Create a new ZPool ===
To create a zfspool, 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-name> <device>
To activate the compression
zfs set compression=lz4 <pool-name>
=== Create a new pool with RAID-0 ===
Minimum 1 Disk
zpool create -f -o ashift=12 <pool-name> <device1> <device2>
=== Create a new pool with RAID-1 ===
Minimum 2 Disks
zpool create -f -o ashift=12 <pool-name> mirror <device1> <device2>
=== Create a new pool with RAID-10 ===
Minimum 4 Disks
zpool create -f -o ashift=12 <pool-name> mirror <device1> <device2> mirror <device3> <device4>
=== Create a new pool with RAIDZ-1 ===
Minimum 3 Disks
zpool create -f -o ashift=12 <pool-name> raidz1 <device1> <device2> <device3>
=== Create a new pool with RAIDZ-2 ===
Minimum 4 Disks
zpool create -f -o ashift=12 <pool-name> raidz2 <device1> <device2> <device3> <device4>
=== 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-name> <device> cache <cache_device>
=== 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-name> <device> log <log_device>
=== Create a new pool with Cache and Log on one Disk ===
It is possible to create ZIL and L2ARC on one SSD. First partition the SSD in 2 partition with parted or gdisk (important: use GPT partition table).
As <device> it is possible to use more devices, like it's shown in "Create a new pool with RAID*".
'''Important: identify device with /dev/disk/by-id/scsi-*<device>''' 
zpool create -f -o ashift=12 <pool-name> <device> log <log_device>
Minimum 1 Disk
zpool create -f -o ashift=12 <pool-name> <device1> <device2>
=== Activate Email notification ===
ZFS >=0.64 has an Email notification daemon.
The daemon send an Email on zfs event like pool errors.
to activate the daemon it is necessary to edit /etc/zfs/zed.d/zed.rc with your favored editor.
<b>Important: the only settings that is required is ZED_EMAIL_ADDR the other are optional.</b>
== Add Cache and Log to existing pool ==
If you have an pool without cache and log.
First partition the SSD in 2 partition with parted or gdisk (important: use GPT partition table).
'''Important: identify device with /dev/disk/by-id/scsi-*<device>''' 
zpool add -f <pool-name> cache <device1.part1> log <device1.part2>
'''Example: Proxmox with zfs as rootFS'''
zpool add -f rpool cache /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_drive-scsi2-part1 \
log /dev/disk/by-id/scsi-0QEMU_QEMU_HARDDISK_drive-scsi2-part2
== Changing a failed Device ==
zpool replace -f <pool-name> <old device> <new-device>
== Using ZFS Storage Plugin (via Proxmox VE GUI or shell) ==
If the zpool is created, you can use it on Proxmox VE GUI and CLI.
=== Adding a ZFS storage via GUI ===
[[Image:Screen-Add-ZFS1.png|thumb]] [[Image:Screen-Add-ZFS2.png|thumb]]
Go to Datacenter/Storage and use the ZFSStorage plugin to add your zpool (select ZFS).
*ID is for identification of the Storage
*the checkbox ZFS Pool shows all existing pools (use CLI to create more)
*Thin provisioning: allocate not all space immediately by creating virtual disks
=== Adding a ZFS storage via CLI ===
To create it by CLI use
pvesm add zfspool <storage-ID> -pool <pool-name>
=== Adding ZFS root file-system as storage with Plugin  ===
If you install Proxmox with ZFS, VM disks on local storage will be simple files, so you will have 2 layers of file system though reducing performance. This is especially true if you choose qcow2 format, because then you will have a "copy on write" image disk (qcow2) that writes on a "copy on write" file system (ZFS).
To avoid it, thanks to the flexibility of ZFS, you can create an additional file system on the default storage pool and, thanks to the ZFS Plugin, use it as "block device" storage have VMs disks created as volumes inside ZFS. In such configuration you keep the ZFS benefit of snapshots and checksuming.
To obtain is do the following:
Create a new filesystem (it's better have a distinct one, separate from the ROOT one), that we will call "zfsdisks" in this example, from shell:
zfs create rpool/zfsdisks
Now add it to the storage (Datacenter -> [Storage] -> Add, choose "ZFS", ad ID let's call it, for example, "zfsvols", as "ZFS Pool" choose "rpool/zfsdisks", set "thin provisioning" and you are ok.
When you create a VM choose "zfsvols" as storage.
cat /etc/pve/storage.cfg
zfspool: zfsvols
        pool rpool/zfsdisks
        content images,rootdir
If you want to add the storage from the shell, you will have the same result (even if /etc/pve/storage.cfg does not exists yet) with
pvesm add zfspool zfsvols -pool rpool/zfsdisks -content images,rootdir -sparse
With a VM 100 and one disk you will have something like that
# zfs list
NAME                          USED  AVAIL  REFER  MOUNTPOINT
rpool                        4.78G  26.0G    96K  /rpool
rpool/ROOT                    676M  26.0G    96K  /rpool/ROOT
rpool/ROOT/pve-1              676M  26.0G  676M  /
rpool/swap                    4.12G  30.1G    64K  -
rpool/zfsdisks                160K  26.0G    96K  /rpool/zfsdisks
rpool/zfsdisks/vm-100-disk-1    64K  26.0G    64K  -
so you see that it's a "block device" and thin provisioning is used (see 'USED' column that states 64K)
== Limit ZFS memory usage ==
It is good to use max 50 percent (which is default) of the system memory for ZFS arc to prevent performance shortage of the host.
Use your preferred editor to change the config in /etc/modprobe.d/zfs.conf and insert:
options zfs zfs_arc_max=4299967296
This example setting limits the usage to 4GB.
'''IMPORTANT: If your root fs is ZFS you must update your initramfs every time this value changes.'''
update-initramfs -u
If your Server has enough memory (i.e. you do not over commit memory) and you have a SWAP partition on ZFS lower the swappiness value to avoid problems and hang ups, a good value for servers is 10:
sysctl -w vm.swappiness=10
To make the '''swappiness persistence''' open "/etc/sysctl.conf" with an editor of your choice
and add the following line.
vm.swappiness = 10
== Misc ==
=== QEMU tuning ===
see thread on proxmox forum, per user Nemesiz:
zfs set primarycache=all tank
*kvm config:
* change cache to Write Back
:You can do it using web GUI or manually. Example:
ide0: data_zfs:100/vm-100-disk-1.raw,cache=writeback
if not set this happened:
qm start 4016
kvm: -drive file=/data/pve-storage/images/4016/vm-4016-disk-1.raw,if=none,id=drive-virtio1,aio=native,cache=none: could not open disk image /data/pve-storage/images/4016/vm-4016-disk-1.raw: Invalid argument
(On PVE 4.x, you may also see warnings about the filesystem not supporting O_DIRECT.)
=== Example configurations for running Proxmox VE with ZFS ===
==== Install on a high performance system ====
As of 2013 and later, high performance servers have 16-64 cores, 256GB-1TB RAM and potentially many 2.5" disks and/or a PCIe based SSD with half a million IOPS. High performance systems benefit from a number of custom settings, for example enabling compression typically improves performance.
* If you have a good number of disks keep organized by using aliases. Edit /etc/zfs/vdev_id.conf to prepare aliases for disk devices found in /dev/disk/by-id/ :
# run 'udevadm trigger' after updating this file
alias a0        scsi-36848f690e856b10018cdf39854055206
alias b0        scsi-36848f690e856b10018cdf3ce573fdeb6
alias a1        scsi-36848f690e856b10018cdf40f5b277cbc
alias b1        scsi-36848f690e856b10018cdf43a5db1b99b
alias a2        scsi-36848f690e856b10018cdf4575f652ad0
alias b2        scsi-36848f690e856b10018cdf47761587cec
Use flash for caching/logs. If you have only one SSD, use parted of gdisk to create a small partition for the ZIL (ZFS intent log) and a larger one for the L2ARC (ZFS read cache on disk). Make sure that the ZIL is on the first partition. In our case we have a Express Flash PCIe SSD with 175GB capacity and setup a ZIL with 25GB and a L2ARC cache partition of 150GB.
*edit /etc/modprobe.d/zfs.conf to apply several tuning options for high performance servers:
# ZFS tuning for a proxmox machine that reserves 64GB for ZFS
# Don't let ZFS use less than 4GB and more than 64GB
options zfs zfs_arc_min=4294967296
options zfs zfs_arc_max=68719476736
# disabling prefetch is no longer required
options zfs l2arc_noprefetch=0
*create a zpool of striped mirrors (equivalent to RAID10) with log device and cache and always enable compression:
zpool create -o compression=on -f tank mirror a0 b0 mirror a1 b1 mirror a2 b2 log /dev/rssda1 cache /dev/rssda2
*check the status of the newly created pool:
root@proxmox:/# zpool status
  pool: tank
state: ONLINE
  scan: none requested
        tank        ONLINE      0    0    0
          mirror-0  ONLINE      0    0    0
            a0      ONLINE      0    0    0
            b0      ONLINE      0    0    0
          mirror-1  ONLINE      0    0    0
            a1      ONLINE      0    0    0
            b1      ONLINE      0    0    0
          mirror-2  ONLINE      0    0    0
            a2      ONLINE      0    0    0
            b2      ONLINE      0    0    0
          rssda1    ONLINE      0    0    0
          rssda2    ONLINE      0    0    0
errors: No known data errors
Using PVE 2.3 on a 2013 high performance system with ZFS you can install Windows Server 2012 Datacenter Edition with GUI in just under 4 minutes.
===How to prevent lvm of scanning zvols===
It is not necessary but to reduce confusion, when using lvm it is a good practice. 
open the config file /etc/lvm/lvm.conf
and insert the following line
filter = [ "r|/dev/zd*|" ]
== Troubleshooting and known issues ==
=== ZFS packages are not installed ===
If you upgraded to 3.4 or later, zfsutils package is not installed. You can install it with apt:
apt-get install zfsutils zfs-initramfs
=== Grub boot ZFS problem ===
*Symptoms: stuck at boot with an blinking prompt.
*Reason: If you ZFS raid it could happen that your mainboard does not initial all your disks correctly and Grub will wait for all RAID disk members - and fails. It can happen with more than 2 disks in ZFS RAID configuration - we saw this on some boards with ZFS RAID-0/RAID-10
=== Boot fails and goes into busybox ===
If booting fails with something like
No pool imported. Manually import the root pool
at the command prompt and then exit.
Hint: try: zpool import -R /rpool -N rpool
is because zfs is invoked too soon (it has happen sometime when connecting a SSD for future ZIL configuration). To prevent it there have been some suggestions in the forum.
Try to boot following the suggestions of busybox or searching the forum, and try ONE of the following:
a) edit /etc/default/grub and add "rootdelay=10" at GRUB_CMDLINE_LINUX_DEFAULT (i.e. GRUB_CMDLINE_LINUX_DEFAULT="rootdelay=10 quiet") and then issue a # update-grub
b) edit /etc/default/zfs, set ZFS_INITRD_PRE_MOUNTROOT_SLEEP='4', and then issue a "update-initramfs -k 4.2.6-1-pve -u"
=== ZFS mounting workaround ===
The default ZFS mount -a script runs too late in the boot process for most system scripts. The following helps to mount ZFS correctly. This is only necessary if you do not use ZFS as root file-system and if you use ZFS as an additional directory storage.
2014-01-22: the info below came from this excellent wiki page:  http://wiki.complete.org/ConvertingToZFS
*Edit /etc/default/zfs and set ZFS_MOUNT='yes'
*edit /etc/insserv.conf,
:and at the end of the $local_fs line,
:add zfs-mount (without a plus).
# All local file-systems are mounted (done during boot phase)
$local_fs      +mountall +mountall-bootclean +mountoverflowtmp +umountfs
edit /etc/init.d/zfs-mount and find three lines near the top, changing them like this:
# Required-Start:
# Required-Stop:
# Default-Start: S
''note remove the Required-Start and -Stop entries.''
*Activating init.d changes Then run:
insserv -v -d zfs-mount
I had an issue with pve storage on ZFS, before pve would start before ZFS and create directories at the ZFS mount point. To fix that start single user mode and remove the directories [ make sure  they are empty.... ].
also see https://github.com/zfsonlinux/pkg-zfs/issues/101
=== SWAP on ZFS ===
SWAP on ZFS on Linux may generate some Troubles, like blocking the server or generating a high IO load, often seen when starting a Backup to an external Storage.
To work around this follow [[#Limit ZFS memory usage]] and especially lower the vm.swappiness value.
=== Snapshot of LXC on ZFS ===
If you can't create a snapshot of an LXC container on ZFS and you get following message:
INFO: rsync: set_acl: sys_acl_set_file(archiv, ACL_TYPE_DEFAULT): Operation not supported (95)
you can run following commands 
zfs create -o mountpoint=/mnt/vztmp rpool/vztmp
zfs set acltype=posixacl rpool/vztmp
Now set /mnt/vztmp in your /etc/vzdump.conf for tmp
== Glossary ==
*ZPool is the logical unit of the underlying disks, what zfs use.
*ZVol is an emulated Block Device provided by ZFS
*ZIL is ZFS Intent Log, it is a small block device ZFS uses to write faster
*ARC is Adaptive Replacement Cache and located in Ram, its the Level 1 cache.
*L2ARC is Layer2 Adaptive Replacement Cache and should be on an fast device (like SSD).
== Further readings about ZFS ==
*https://www.freebsd.org/doc/handbook/zfs.html (even if written for freebsd, of course, I found this doc is extremely clear even for less "techie" admins [note by m.ardito])
*https://pthree.org/2012/04/17/install-zfs-on-debian-gnulinux/ (and all other pages linked there)
and this has some very important information to know before implementing zfs on a production system.
Very well written manual pages
man zfs
man zpool
[[Category:HOWTO]] [[Category:Installation]] [[Category:Technology]]
[[Category:HOWTO]] [[Category:Installation]] [[Category:Technology]]
[[Category:Reference Documentation]]
[[Category:Reference Documentation]]

Revision as of 10:47, 23 September 2016

Storage pool type: zfspool

This backend allows you to access local ZFS pools (or ZFS file systems inside such pools).


The backend supports the common storage properties content, nodes, disable, and the following ZFS specific properties:


Select the ZFS pool/filesystem. All allocations are done within that pool.


Set ZFS blocksize parameter.


Use ZFS thin-provisioning. A sparse volume is a volume whose reservation is not equal to the volume size.


The mount point of the ZFS pool/filesystem. Changing this does not affect the mountpoint property of the dataset seen by zfs. Defaults to /<pool>.

Configuration Example (/etc/pve/storage.cfg)
zfspool: vmdata
        pool tank/vmdata
        content rootdir,images

File naming conventions

The backend uses the following naming scheme for VM images:

vm-<VMID>-<NAME>      // normal VM images
base-<VMID>-<NAME>    // template VM image (read-only)
subvol-<VMID>-<NAME>  // subvolumes (ZFS filesystem for containers)

This specifies the owner VM.


This can be an arbitrary name (ascii) without white space. The backend uses disk[N] as default, where [N] is replaced by an integer to make the name unique.

Storage Features

ZFS is probably the most advanced storage type regarding snapshot and cloning. The backend uses ZFS datasets for both VM images (format raw) and container data (format subvol). ZFS properties are inherited from the parent dataset, so you can simply set defaults on the parent dataset.

Table 1. Storage features for backend zfs
Content types Image formats Shared Snapshots Clones

images rootdir

raw subvol





It is recommended to create an extra ZFS file system to store your VM images:

# zfs create tank/vmdata

To enable compression on that newly allocated file system:

# zfs set compression=on tank/vmdata

You can get a list of available ZFS filesystems with:

# pvesm zfsscan