ZFS is often called the last word in file systems.
It is a new approach to deal with large pools of disks originally invented by Sun.
It was later then ported to FreeBSD, MacOS (only 10.5) and Linux.
This text should show some of the basic feature of ZFS and demonstrate them handson by example.
Prerequisites
-> FreeBSD
-> Solaris
-> MacOS (only Userland)
In our example we use
SunOS openindiana 5.11 oi_151a5 i86pc i386 i86pc Solaris.
as an environment.
But most commands also work on the other systems.
Since we do all the work within a VM, our commands have the pattern:
Input VM:
command
Output VM:
result
Pool Creation
The first information we need is the number of disk, present in our environment.
There are several ways to get a basic disk listing. Under (Open-)Solaris this can be done with:
Input VM:
format < /dev/null
Output VM:
AVAILABLE DISK SELECTIONS:
0. c4t0d0
/pci@0,0/pci8086,2829@d/disk@0,0
1. c5t0d0
/pci@0,0/pci1000,8000@16/sd@0,0
2. c5t1d0
/pci@0,0/pci1000,8000@16/sd@1,0
3. c5t2d0
/pci@0,0/pci1000,8000@16/sd@2,0
4. c5t3d0
/pci@0,0/pci1000,8000@16/sd@3,0
5. c5t4d0
/pci@0,0/pci1000,8000@16/sd@4,0
6. c5t5d0
/pci@0,0/pci1000,8000@16/sd@5,0
7. c5t6d0
/pci@0,0/pci1000,8000@16/sd@6,0
8. c5t7d0
/pci@0,0/pci1000,8000@16/sd@7,0
about Pools
With ZFS it is possible to create different kinds of pools on a specific number of disk.
You can also create several pools within one system.
The following Pools are possible and most commonly used:
| Type |
Info |
Performance |
Capacity |
Redundancy |
Command |
| JBOD |
Just a bunch of disks. In theory it is possible to create on pool for each disk in the system, although this is not quite commonly used. |
of each disk |
of each disk |
– |
zpool create disk1 pool1
zpool create disk1 pool2
… |
| Stripe |
This is equivalent to RAID0, the data is distributed over all disks in the pool. If one disks fails, all the data is lost. But you can also stripe several Pools (e.g. two raidz pools) to have better redundancy. |
very high |
N Disks |
no |
zpool create disk1 disk2 pool1 |
| Mirror |
This is equivalent to RAID1, the data is written to both disks in the Pool. Restoring a pool (resilvering) is less efficient, since the data needs to be copied from the remaining disk. |
normal |
N-1 Disks |
+1 |
zpool create mirror disk1 disk2 pool1 |
| Raidz |
This is equivalent to RAID5. One disk contains the parity data. Restoring a pool (resilvering) is less efficient, since the data needs to be copied from the remaining disks. |
high |
N-1 Disks |
+1 |
zpool create raidz disk1 disk2 disk3 pool1 |
| Raidz2 |
This is equivalent to RAID6. Two disks containing the parity data. Restoring a pool (resilvering) is less efficient, since the data needs to be copied from the remaining disk with parity data. |
high |
N-2 Disks |
+2 |
zpool create raidz2 disk1 disk2 disk3 disk4 pool1 |
| Raidz3 |
There is no real equivalent existing for that one. You have basically three disks with parity data. |
high |
N-3 Disks |
+3 |
zpool create raidz3 disk1 disk2 disk3 disk4 disk5 pool1 |
You can also add hot-spares for a better fallback behaviour, SSDs for caching reads (cache) and writes (logs).
I also created a benchmark with various combinations.
create a basic Pool (raidz)
Input VM:
zpool create tank raidz c5t0d0 c5t1d0 c5t2d0
...
zpool status
Output VM:
pool: tank
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
tank ONLINE 0 0 0
raidz1-0 ONLINE 0 0 0
c5t0d0 ONLINE 0 0 0
c5t1d0 ONLINE 0 0 0
c5t2d0 ONLINE 0 0 0
errors: No known data errors
(Raid5)
You already can access the newly created pool:
Input VM:
ls -al /tank
Output VM:
...
total 4
drwxr-xr-x 2 root root 2 2012-10-23 22:02 .
drwxr-xr-x 25 root root 28 2012-10-23 22:02 ..
create a basic Pool (raidz) with one spare drive
Input VM:
zpool create tank raidz1 c5t0d0 c5t1d0 c5t2d0 spare c5t3d0
...
zpool status
Output VM:
pool: tank
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
tank ONLINE 0 0 0
raidz1-0 ONLINE 0 0 0
c5t0d0 ONLINE 0 0 0
c5t1d0 ONLINE 0 0 0
c5t2d0 ONLINE 0 0 0
spares
c5t3d0 AVAIL
errors: No known data errors
List the availibe Layout
Input VM:
zpool list
Output VM:
NAME SIZE ALLOC FREE EXPANDSZ CAP DEDUP HEALTH ALTROOT
tank 1,46G 185K 1,46G - 0% 1.00x ONLINE -
*The 1,5G does not reflect the real availible space. If you copy a 1G File to the Pool it will use 1,5G (1G + 512M Parity).
create a stripped pool
Input VM:
zpool create tank raidz1 c5t0d0 c5t1d0 c5t2d0 raidz1 c5t4d0 c5t5d0 c5t6d0
Output VM:
pool: tank
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
tank ONLINE 0 0 0
raidz1-0 ONLINE 0 0 0
c5t0d0 ONLINE 0 0 0
c5t1d0 ONLINE 0 0 0
c5t2d0 ONLINE 0 0 0
raidz1-1 ONLINE 0 0 0
c5t4d0 ONLINE 0 0 0
c5t5d0 ONLINE 0 0 0
c5t6d0 ONLINE 0 0 0
errors: No known data errors
(Raid50 = Raid5 + Raid5)
deal with disk failures
Input VM:
zpool create tank raidz1 c5t0d0 c5t1d0 c5t2d0 spare c5t3d0
Failure Handling
Input Host:
echo /dev/random >> 1.vdi
Wait for it…
or Input VM:
pool: tank
state: DEGRADED
status: One or more devices has experienced an unrecoverable error. An
attempt was made to correct the error. Applications are unaffected.
action: Determine if the device needs to be replaced, and clear the errors
using 'zpool clear' or replace the device with 'zpool replace'.
see: http://illumos.org/msg/ZFS-8000-9P
scan: resilvered 66K in 0h0m with 0 errors on Tue Oct 23 22:14:19 2012
config:
NAME STATE READ WRITE CKSUM
tank DEGRADED 0 0 0
raidz1-0 DEGRADED 0 0 0
spare-0 DEGRADED 0 0 0
c5t0d0 DEGRADED 0 0 64 too many errors
c5t3d0 ONLINE 0 0 0
c5t1d0 ONLINE 0 0 0
c5t2d0 ONLINE 0 0 0
spares
c5t3d0 INUSE currently in use
errors: No known data errors
Input VM:
zpool clear tank
...
zpool detach tank c5t0d0
zpool replace tank c5t0d0 c5t7d0
Create File systems
Input VM:
zfs create tank/home
zfs create tank/home/user1
...
chown -R user:staff /tank/home/user1
...
zfs get all tank/home/user1
...
zfs set sharesmb=on tank/home/user1
...
zfs set quota=500M tank/home/user1
Copy File from MacOS into SMB Share.
Snapshot
Input VM:
zfs snapshot tank/home/user1@basic
...
zfs list
...
zfs list -t snapshot
Output VM:
NAME USED AVAIL REFER MOUNTPOINT
rpool1/ROOT/openindiana@install 84,0M - 1,55G -
tank/home/user1@basic 0 - 42,6K -
Input VM:
zfs snapshot -r tank/home@backup
...
zfs list -t snapshot
Output VM:
NAME USED AVAIL REFER MOUNTPOINT
rpool1/ROOT/openindiana@install 84,0M - 1,55G -
tank/home@backup 0 - 41,3K -
tank/home/user1@basic 0 - 42,6K -
tank/home/user1@backup 0 - 42,6K -
Input VM:
zfs clone tank/home/user1@basic tank/home/user2
Output VM:
tank/home/user2 1,33K 894M 70,3M /tank/home/user2
Restoring Snapshots
Delete ZIP File in SMB-Share.
Input VM:
ls -al tank/home/user1
...
zfs rollback tank/home/user1@backup
Output VM:
ls -al tank/home/user1
Resizing a Pool
Input VM:
zpool list
...
zpool replace tank c5t0d0 c5t4d0
zpool replace tank c5t1d0 c5t5d0
zpool replace tank c5t2d0 c5t6d0
...
zpool scrub tank
...
zpool list
Output VM:
NAME SIZE ALLOC FREE EXPANDSZ CAP DEDUP HEALTH ALTROOT
tank 1,46G 381K 1,46G 1,50G 0% 1.00x ONLINE -
Input VM:
zpool set autoexpand=on tank
Using ZFS for Backups
Bash-Script
rsync -avrz --progress --delete /Users/user root@nas.local::user-backup/
backupdate=$(date "+%Y-%m-%d")
ssh root@nas.local zfs snapshot tank/backup@$backupdate
