ZFS

From CaveLab

ZFS is a combined file system and logical volume manager designed by Sun Microsystems. ZFS is scalable, and includes extensive protection against data corruption, support for high storage capacities, efficient data compression, integration of the concepts of filesystem and volume management, snapshots and copy-on-write clones, continuous integrity checking and automatic repair, RAID-Z, native NFSv4 ACLs, and can be very precisely configured. (source: Wikipedia)

Pool

RAID options

Description Type Behavior Multiple VDEVs in single pool
Striped Single disk Striped VDEVs, like RAID0
Mirrored mirror Like RAID1 Striped mirror, like RAID10
RAIDZ raidz Like RAID5 Nested RAIDZ, like RAID50/60
RAIDZ2 raidz2 Like RAID6
RAIDZ3 raidz3

Create

sudo zpool create testpool -o ashift=12 raidz1 /dev/disk/by-id/ata-ST2000DM001-1CH164_W1E5ETF9 /dev/disk/by-id/ata-ST2000DM001-1CH164_Z1E68GLR /dev/disk/by-id/ata-ST2000DM001-1CH164_Z1E6CQPW

or

sudo zfs create testpool -o ashift=12 -o mountpoint=/srv/testpool raidz1 /dev/disk/by-id/ata-ST2000DM001-1CH164_W1E5ETF9 /dev/disk/by-id/ata-ST2000DM001-1CH164_Z1E68GLR /dev/disk/by-id/ata-ST2000DM001-1CH164_Z1E6CQPW
  • -o ashift=12 set to 4K sectors
  • -o mountpoint=/srv/testpool set custom mount point
Make sure that you create your pools such that the vdevs have the correct alignment shift for your storage device’s size. if dealing with flash media, this is going to be either 12 (4K sectors) or 13 (8K sectors). http://open-zfs.org/wiki/Performance_tuning
sudo zpool status

Add VDEV

sudo zpool add (-n) (-f) tank0 -o ashift=12 raidz2 /dev/disk1 /dev/disk2 /dev/disk3 /dev/disk4
  • -n dry-run, shows what the new configuration will look like
  • -f required if the disks are of different sizes

Compression

sudo zfs set compression=lz4 testpool
sudo zfs get compression testpool
Size reduced with compression:
  • VirtualBox VM: 48G → 34G
  • Documents, photos, videos, etc: 395G → 388G

Mount point

sudo zfs set mountpoint=/srv/testpool testpool
zfs get mountpoint

Stats

zpool iostat testpool 1
zfs get space (pool/dataset)

Scrub

zpool scrub files

Replace disk

  • Offline the disk, if necessary, with the zpool offline command.
  • Remove the disk to be replaced.
  • Insert the replacement disk.
  • Run the zpool replace command.
zpool offline testpool disk-old-id
zpool replace (-f) testpool disk-old-id /dev/disk/by-id/new
  • -f forces use of new disk, even if its appears to be in use

Destroy

sudo zpool destroy tank

Dataset

# Create
sudo zfs create testpool/test

# Remove
sudo zfs destroy testpool/test (-r)

# List
sudo zfs list

# Rename
sudo zfs rename testpool/test testpool/new_test

Snapshot

Snapshots aren't pointers, they are time stamps.

ZFS is a tree of blocks and each block has a set of metadata, one of which is its birth date. Any block with a birth date older than the snapshot time stamp is part of that snapshot.

What happens is something like this; your 10 MB file gets written to 79 128k blocks and you take a snapshot. You then change 5 MB of the file causing 40 new blocks to be written. Then ZFS will try and remove the outdated 40 blocks and while its doing that its checking to see if any block is part of a snapshot, leaving them alone if they are. Thus you now have a "snapshot" worth 5 MB. [1]

# Create
sudo zfs snap testpool/test@snap

# Destroy
sudo zfs destroy testpool/test@snap

# List
sudo zfs list -t snapshot

# Clone
sudo zfs clone testpool/test@snap testpool/test/snap-clone

Send and receive

# local
zfs send tank/data@snap1 | zfs recv spool/ds01

# allow regular user to send
sudo zfs allow username send,snapshot,mount tank/data
# remote
zfs send tank/dana@snap1 | ssh host2 zfs recv spool/ds01

# allow regular user to receive
sudo zfs allow username receive,create,mount spool
Delegating mount rights to a regular user doesn't work in Linux...

Resources

References