KfWiki: Linux File Systems

Linux File Systems

SubTopics

LVM
Encrypted FileSystem
NFS
FileSystems Benchmark
Resize file system
Quota
ISCSI
Zfs on Linux
BTRFS
OCFS2
GlusterFS
Striping
EFI GPT
Luks Encrypted Filesystem

Other things

Filesystem and UUID - http://linuxshellaccount.blogspot.com/2008/08/how-to-manage-your-disk-by-uuid-on.html

undelete

Foremost http://www.ubuntu-unleashed.com/2008/04/howtorecover-and-undelete-text-file-in.html

hd tuning

# Query your hdd for maximum multi sector transfer supported.
hdparm -i /dev/hda | grep MaxMultiSect

# setting 32-bit IO_support flag to 1
# setting multcount to 16
hdparm -c1 -k1 -m16 /dev/hda

# Benchmark using hdparm
hdparm -tT /dev/hda

# Tunning LVM
# Add to /etc/sysctl.conf
vm.max-readahead=256
vm.min-readahead=128

Mount options

File system	option
ext2	notail,noatime
ext3	data=journal,errors=remount-ro
ext4dev	extents (tune2fs -O extents or mke2fs -O extents before mounting with this option)

IO tuning reference: http://www-128.ibm.com/developerworks/linux/library/l-fs8.html

tmpfs

TMPFS is a Linux file system which keeps all files in virtual memory. It is supported by the Linux kernel from version 2.4 and up. TMPFS distinguishes itself from ramdisk by using memory dynamically, through swapping unneeded pages onto swap space.

dd if=/dev/zero of=/.tmpfs bs=1024 count=250000

mkfs -t ext3 /.tmpfs

mount -o loop,noexec,nosuid,rw /.tmpfs /tmp

chmod 0777 /tmp

chmod +t /tmp

To make this permanent, add the following to /etc/fstab

/.tmpfs /tmp ext3 loop,rw,nosuid,noexec 0 0

loop mount

Take a look at the command losetup

swap file

Sometimes it's useful to add swap space using swap files. To create a 500m swap file:

dd if=/dev/zero of=swap_file_01 bs=1024 count=500000

mkswap swap_file_01

swapon swap_file_01

To make this permanent, add this to /etc/fstab

/swap_file_01 swap swap defaults 0 0

It's a little more work to do the same thing under freebsd 4.x

# Create a vnode

cd /dev

sh MAKEDEV vn0

# Create a swapfile (/usr/swap0):

dd if=/dev/zero of=/usr/swap0 bs=1024k count=1024

# Set proper permissions on (/usr/swap0):

chmod 0600 /usr/swap0

#Enable the swap file in /etc/rc.conf:

swapfile="/usr/swap0"   # Set to name of swapfile if aux swapfile desired.

# Reboot the machine or to enable the swap file immediately, type:

vnconfig -e /dev/vn0b /usr/swap0 swap

# Check swap usage

swapinfo

It's yet a little different on FreeBSD 5.x

#Create a swapfile (/usr/swap0):

dd if=/dev/zero of=/usr/swap0 bs=1024k count=64

#Set proper permissions on (/usr/swap0):

chmod 0600 /usr/swap0

#Enable the swap file in /etc/rc.conf:

swapfile="/usr/swap0"   # Set to name of swapfile if aux swapfile desired.

#Reboot the machine or to enable the swap file immediately, type:

mdconfig -a -t vnode -f /usr/swap0 -u 0 && swapon /dev/md0

# Check swap usage

swapinfo

Raw partition

On Redhat, raw disk can be set up by

raw /dev/raw/raw1 /dev/vg0/lv_raw

Then add the corresponding entry to /etc/sysconfig/rawdevices, and you're done!

/dev/raw/raw1 /dev/vg0/lv_raw

raw has been dropped on the Fedora / CentOS5 kernel. They said raw devices are now being created via udev http://kbase.redhat.com/faq/FAQ_105_9616.shtm followed by udevcontrol reload_rules But I got this error -

Jan 14 21:05:30 homme udevd[555]: add_to_rules: invalid ACTION operation
Jan 14 21:05:30 homme udevd[555]: add_to_rules: invalid rule '/etc/udev/rules.d/90-raw.rules:1'

Raw device with loopback mount: see http://www.osdev.org/osfaq2/index.php/Disk%20Images%20Under%20Linux

Hard link vs Soft link

A link provides multiple access point to the same file.
A soft link is a pointer to the real file. One can identify a soft link by ls -l and the first column should be displayed as lrwxrwxrwx. When the real file is removed, the soft links will be broken.
A hard link is a file pointing to the same inode number as the original file. To identify hard links, do a ls -i and compare the inode number. Alternatively, one can use find / -inum 1234 to locate all the hard links. One can think of the original file as the first hard link to the real file. Until the last hard link is deleted, all remaining hard links are still pointing to the real file.

Fixing ext3 journals

If this is observed from syslog and your fs is automatically mounted as read-only

EXT3-fs error (device hda1) in start_transaction: Journal has aborted

Remove the journal, do an fsck, and rebuild the journal.

# tune2fs -O ^has_journal /dev/hda1
# fsck -yC /dev/hda1
# tune2fs -j /dev/hda3

Disk labeling

# show all labels
/sbin/blkid

# label a disk
e2fslabel /dev/hdxn

Determine a partition's filesystem

head -c 1m /dev/sdc2 | file -

Copying existing partition to a new drive

I use tar. If it's the same filesystem, one can use dd or dump/restore. Check tar's man page, it seems the exclude option is slightly different on different distro. On CentOS, you don't need a = sign. On Debian, you do !

tar cvfS - / --exclude /tmp --exclude /sys --exclude /dev --exclude /mnt  | (cd /mnt/newdrive; tar xfS -)
# a variant if the target is remote -
tar cfS - / --exclude /tmp --exclude /sys --exclude /dev --exclude /mnt | ssh root@1.2.3.4 (cd / ; tar xvfS -)

Seems cpio is cool too

find / -xdev | cpio -pvmd /mnt

Or use dump | restore

cd /newdrive
dump -0uan -f - / | restore -r -f -

Or use dd

dd if=/dev/sda1 of=/dev/sdb1 bs=4k

If it's just one file

gzip -c bigfile.log | ssh me@blah.domain.tld zcat ">" /tmp/bigfile.log

Identify processes on current file

fuser $filename

Ram drive

none                    /memtmp                 tmpfs   size=3G         0 0

Re-read partition table

hdparm -z
or
partprobe

Resize reiserfs

umount /dev/mapper/vg0-reiserfs
fsck -fC /dev/mapper/vg0-reiserfs
resize_reiserfs -s 100G /dev/mapper/vg0-reiserfs
lvresize --size 100G  /dev/mapper/vg0-reiserfs
sync
# show free space
vgdisplay

Remove raid

mdadm --manage --fail /dev/md0
mdadm --manage  --remove /dev/md0
mdadm --manage --stop /dev/md0