Linux/Btrfs

Btrfs

http://btrfs.wiki.kernel.org

Installation

# centos
yum install btrfs-progs
# WARNING! - Btrfs Btrfs v0.20-rc1 IS EXPERIMENTAL

# ubuntu:
apt-get install btrfs-tools
# WARNING! - Btrfs v3.12 IS EXPERIMENTAL

Test JBOD Failure Recovery

# create image file
dd if=/dev/urandom of=img0 bs=1M count=20
dd if=/dev/urandom of=img1 bs=1M count=20
dd if=/dev/urandom of=img2 bs=1M count=20

# create loopback devices
losetup /dev/loop0 img0
losetup /dev/loop1 img1
losetup /dev/loop2 img2
# show loop back devices
losetup -a

# verify btrfs module is loaded
lsmod | grep btrfs

# make btrfs file system
mkfs.btrfs -L "btrfs-test" /dev/loop0 /dev/loop1 /dev/loop2

# btrfs device scan
btrfs device scan
btrfs filesystem show  # should show the devices

# see that the block id label is picked up
blkid | grep btrfs-test

# mount btrfs file system
mount /dev/loop0 /mnt

#traffic
cd /mnt
for i in 1 2 3 4 5 ; do mkdir $i ; cd $i ; dd if=/dev/urandom of=$i.dat bs=1M count=20 ; md5sum $i.dat> $i.dat.md5 ; cd .. ; done
cd ~

# umount btrfs file system
umount /mnt

# simulate destroying one drive
#losetup -d /dev/loop0
losetup -d /dev/loop1

# btrfs device scan
btrfs device scan
btrfs filesystem show  # should show the devices

# try mounting remaining device
#mount /dev/loop1 /mnt  # <-- fails - expected
#mount /dev/loop1 /mnt -oro,degraded  # <-- fails!
#mount /dev/loop1 /mnt -oro,degraded,recovery  # <-- fails!
mount /dev/loop2 /mnt -oro,degraded

# verify data
cd /mnt
for i in 1 2 3 4 5 ; do cd $i ; echo "$i:" ; md5sum -c $i.dat.md5 ; cd .. ; done

# cleanup
umount /mnt  # should already be done
losetup -d /dev/loop0  # should already be done
losetup -d /dev/loop1
losetup -d /dev/loop2
rm -f img0 img1 img2

Note: Failing the first drive fails the to mount at all, but failing a different drive seems to work.

Test mentioned here: https://btrfs.wiki.kernel.org/index.php/Using_Btrfs_with_Multiple_Devices#Filesystem_creation