setup-storage - automatically prepare storage devices


setup-storage [-X] [-f filename] [-d] [-h] [-s] [-D disks] [-L directory]


Using FAI disk_config files, setup-storage Computes effective partition and volume sizes and executes the necessary commands to configure storage devices. It manages disk drives, RAID and LVM volumes, along with encryption and tmpfs. Disks and partitions are designated by disk1.2, disk3.1 etc. and may thus be referenced as such. Currently, setup-storage handles following filesystems: ext2/3/4, vFAT(FAT32), msdos(FAT16), reiserFS, XFS and BTRFS but could easily be extended to further types as well. Once the storage devices are prepared, an appropriate fstab(5) (and possibly also a crypttab(5) file is generated.

Without the -X parameter setup-storage runs in test-only mode and does not execute commands other than writing disk labels to a blank disk.

The exit code of setup-storage is 0 if all operations were performed successfully and non-zero if an error occurs.



Really write the configuration to disk. Otherwise setup-storage runs in test-only mode.

-f filename

Normally setup-storage selects an appropriate configuration from $FAI/disk_config/ by picking the class with the highest priority from classes that has an existing file. If however -f is given the configuration in filename is used.


Enable debugging output. Equivalent to environment variable debug set to a non-zero value. See below for further details.


Perform syntax check of disk_config file only and exit.

-D disks

Specify the list of disk drives to be configured using setup-storage. Overrides the variable disklist. Do not forget to quote this space separated list.

-L directory

Use the specified directory instead of LOGDIR.


Display the synopsis and version info and exit.


Print disk variables as YAML file into disk_var.yml


setup-storage will use the following environment variables:

If option -D is not used, the disklist variable may contain a space separated list of disk drives available in the system. Their order matters as they may be referred to as disk1, and so on, in disk_config. If unset, fai-disk-info will be called to determine the list.


If debug is set to a non-zero value all actions and details to track the operation of setup-storage are printed to stderr.


The location of the config space to find the disk_config directory.


The list of FAI classes to determine the appropriate configuration to choose.


setup-storage generates, fstab, and possibly crypttab (see below) in this directory. Defaults to /tmp/fai if unset. Option -L overrides this.


This variable determines if partitions should be preserved when they are tagged with preserve_reinstall (see below for details). Normally set by the list of FAI flags (FAI_FLAGS).


This variable may contain a list of volume groups, that are ignored by setup-storage. The list can be comma or space separated. It’s used by fai-diskimage.


If setup-storage executes successfully, an fstab(5) file matching the specified configuration is generated as $LOGDIR/fstab. Furthermore the file $LOGDIR/ is generated. This file defines the following variables, if not yet set: SWAPLIST, ROOT_PARTITION, BOOT_PARTITION (which is only set in case this resides on a disk drive), BOOT_DEVICE and PHYSICAL_BOOT_DEVICES (which contains the list of all physical devices having a bootable partition). Both BOOT_PARTITION and BOOT_DEVICE describe the partition and disk/RAID/LVM device hosting the mount point for /boot. If /boot has no extra mount point, / is used instead. PHYSICAL_BOOT_DEVICES can be used to determine where a bootloader should be installed (this is useful if / is on a LVM or RAID device). You may source $LOGDIR/ to get the variables set. The example config space shipped with FAI sources this file in scripts/GRUB_PC/10-setup. If encryption was configured, a proper crypttab(5) file plus key files will be generated.


setup-storage configures storage devices according to a FAI disk_config file. The full grammar describing the syntax of these disk_config files is given below. First we show a number of examples to give an intuition what these should look like.
Simple configuration of /dev/hda

disk_config hda preserve_always:6,7 disklabel:msdos bootable:3

primary /boot 20-100 ext3 rw
primary swap 1000 swap sw
primary / 12000 ext3 rw createopts="-b 2048"
logical /tmp 1000 ext3 rw,nosuid
logical /usr 5000 ext3 rw
logical /var 10%- ext3 rw
logical /nobackup 0- xfs rw

Preserve /dev/hda6 and /dev/hda7. The disklabel is msdos which is the default for x86. Furthermore, partition /dev/hda3 is made bootable.

Create a primary partition /dev/hda1 with a size between 20 and 100 MiB and mount it read-write as /boot; it is formatted using ext3 filesystem.

/dev/hda2 will be a swap space of 1000 MiB

/dev/hda3 should be formatted using ext3 filesystem; when calling mkfs.ext3 the option "-b 2048" is appended.

Create the logical partition /dev/hda5

Make /dev/hda7 at least 10% of the disk size

Use mkfs.xfs to format the partition 8

Create a softRAID

Image grohtml-251251.png

Create 2 partitions of size 20 and 4 GiB, respectively, on disks /dev/sda and /dev/sdb. No file system or fstab entry will be created.

Create a RAID-1 on /dev/sda1 and /dev/sdb1, format using mkfs.ext3 and mount it as /

Create a RAID-1 on /dev/sda2 and /dev/sdd2, prepare as swap space and use it as such later on.

Advanced softRAID features

Image grohtml-251252.png

Create a RAID-1 on /dev/sda1 and /dev/sdd1, format using mkfs.ext2 and mount it as /

Create a RAID-0 on the second partition of the second disk, /dev/sdc1, and /dev/sde1 as a spare partition. The latter may be missing.

The configurations for /dev/sda, /dev/sdd, disk2, /dev/sdc, and /dev/sde are omitted in this example. These devices can be configured similar to the configuration shown in the first example. They may also already be properly set up and hence the configuration can be left out.

Simple LVM example

disk_config sda bootable:1
primary /boot 500 ext3 rw
primary - 4096- - -

disk_config lvm
vg my_pv sda2
my_pv_swap swap 2048 swap sw
my_pv_root / 2048 ext3 rw

Configure /dev/sda with two partitions.

The second of those, /dev/sda2, is then used in the LVM volume group my_pv.

This volume group hosts two logical volumes: _swap and _root.

LVM on software RAID

disk_config disk1
primary - 350 - -
primary swap 2G swap sw,pri=1
primary - 0- - -

disk_config sdb sameas:disk1

disk_config raid fstabkey:uuid
raid1 /boot disk1.1,disk2.1 ext4 rw,noatime,errors=remount-ro
raid1 - disk1.3,disk2.3 - -

disk_config lvm fstabkey:uuid
vg vg_system md1
vg_system-root / 8G ext4 rw,noatime
vg_system-var /var 20G ext4 rw,noatime
vg_system-home /home 10G ext4 rw,noatime,nosuid,nodev
vg_system-tmp /tmp 30G ext4 rw,noatime,nosuid,nodev

Configure the first and second disk identical with three partitions each.

The first partitions of each device are bundled into a RAID 1 (/dev/md0), which will be mounted at /boot.

The third partitions of each device are combined as another RAID 1, but not mounted. Instead, the resulting device /dev/md1 is used to host the LVM volume group vg_system.

Do not forget to install the packages mdadm and lvm2 into a system using this partition scheme. Also set when using dracut as initrd generator.

Crypt example

disk_config /dev/sdb
primary / 21750 ext3 defaults,errors=remount-ro
primary /boot 250 ext3 defaults
logical - 4000 - -
logical - 2000 - -
logical - 10- - -

disk_config cryptsetup
swap swap /dev/sdb5 swap defaults
tmp /tmp /dev/sdb6 ext2 defaults
luks /local00 /dev/sdb7 ext3 defaults,errors=remount-ro createopts="-m 0"

Configure /dev/sdb with 2 primary partitions, one extended partition, and 3 logical partitions.

Encrypt the swap space, /tmp, and /local00. As described in the CAVEATS section, the encryption keys will be stored in a temporary directory only.

Do not forget to install the cryptsetup package on a system using thus scheme.

The initial password for the encrypted device can also be set in the configuration file by using the syntax - luks:"secret"

With a working RAID+LVM configuration, an encryption layer can be added between the RAID and LVM device layers by adding the following cryptsetup configuration. In this case, the encrypted device will be called ’crypt_format_md1’ and will be used as the underlying physical device (PV) in LVM.

disk_config cryptsetup
luks - /dev/md1 - -

Plain disk + LUKS + LVM2 example

disk_config disk1 disklabel:gpt bootable:1 fstabkey:uuid align-at:1M
primary - 100% - -

disk_config cryptsetup
luks:"passwd" - disk1.1 - -

disk_config lvm fstabkey:uuid
vg vg1 disk1.1
vg1-root / 30%- ext4 defaults,errors=remount-ro,noatime,rw

tmpfs example

disk_config tmpfs
tmpfs /tmp RAM:20% defaults
tmpfs /scratch 3GiB defaults

Mount a tmpfs on /tmp with a maximum size equal to 20% of the total amount of RAM in the machine. This is equivalent to using size=20% in the tmpfs mount options.

Mount a tmpfs on /scratch with a maximum size of 3 GiB.

nfs example

disk_config nfs



A diskless client setup. Mount the whole OS file system via NFS from an NFS server onto /.

Simple BTRFS example

disk_config disk1
primary /boot 500 ext4 rw
primary - 2G- - -

disk_config disk2
primary - 2G- - -

disk_config disk3 sameas:disk2
disk_config disk4 sameas:disk2

disk_config btrfs fstabkey:uuid
btrfs raid1 / disk1.2,disk2.1 noatime,subvol=@/
btrfs raid1 /home disk3.1,disk4.1 subvol=@home,noatime

Four disks are used to create a BTRFS RAID. The first disk contains the boot partition, the second partition of disk one and the second disk in its entirety are used to create the / RAID. The third and fourth disks are used to create the /home RAID.

Note that each BTRFS RAID must contain an initial subvolume. This is necessary to use advanced BTRFS features such as snapshots. The initial subvolume name is taken from the subvol mountoption. In the above example those would be @/ and @home.

Subvolume names begin with an @ by convention.

Every BTRFS line must begin with btrfs followed by the RAID-level of the actual data -- NOT metadata! By default metadata uses RAID1, however this can be changed using createopts.

External log device example

disk_config /dev/sda fstabkey:uuid bootable:2
primary / 20GiB ext3 defaults
primary /boot 250 ext2 defaults
primary swap 4GiB swap defaults
logical - 256 ext3_journal -
logical - 256 ext4_journal -
logical - 256 xfs_journal -

disk_config /dev/sdb fstabkey:uuid
primary /mnt/ext3 33% ext3:journal=/dev/sda5 defaults
primary /mnt/ext4 33% ext4:journal=/dev/sda6 defaults
primary /mnt/xfs 33% xfs:journal=/dev/sda7 defaults

Mount an ext3 filesystem on /dev/sdb1 with an external journal on /dev/sda5

Mount an ext4 filesystem on /dev/sdb2 with an external journal on /dev/sda6

Mount an XFS filesystem on /dev/sdb3 using /dev/sda7 as the log device

Example using a GPT partition table and EFI/ESP partition

disk_config disk1 disklabel:gpt fstabkey:uuid bootable:1
primary /boot/efi 200 vfat rw
primary / 1G- ext4 rw

Creating a ESP partition is very simple. Use a GPT partition table, create a vfat partition which is also bootable. That’s it.


This section describes the syntax of disk_config files

file ::= <lines> EOF

lines ::= EOL
/* empty lines or whitespace only */
| <comment> EOL
| <config> EOL

comment ::= #.*

config ::= disk_config lvm( <lvmoption>)*
| disk_config raid( <raidoption>)*
| disk_config cryptsetup( <cryptsetupoption>)*
| disk_config tmpfs
| disk_config end
| disk_config disk[[:digit:]]+( <option>)*
| disk_config [^[:space:]]+( <option>)*
/* fully qualified device-path or short form, like hda, whereby full
* path is assumed to be /dev/hda; may contain shell globbing such
* as /dev/disk/by-id/scsi-* */
| <volume>

lvmoption ::= /* empty */
| preserve_always:([^/,\s\-]+-[^/,\s\-]+(,[^/,\s\-]+-[^/,\s\-]+)*|all)
/* preserve volumes -- always */
| preserve_reinstall:([^/,\s\-]+-[^/,\s\-]+(,[^/,\s\-]+-[^/,\s\-]+)*|all)
/* preserve volumes -- unless the system is installed for the
first time */
| preserve_lazy:([^/,\s\-]+-[^/,\s\-]+(,[^/,\s\-]+-[^/,\s\-]+)*|all)
/* preserve volumes -- unless these don’t exist yet */
| always_format:([^/,\s\-]+-[^/,\s\-]+(,[^/,\s\-]+-[^/,\s\-]+)*|all)
/* run mkfs on the volumes, even if marked as preserve */
| resize:([^/,\s\-]+-[^/,\s\-]+(,[^/,\s\-]+-[^/,\s\-]+)*|all)
/* attempt to resize partitions */
| fstabkey:(device|label|uuid)
/* when creating the fstab, the key used for defining the device
may be the device (/dev/xxx), a label given using -L, or the uuid

raidoption ::= /* empty */
| preserve_always:([[:digit:]]+(,[[:digit:]]+)*|all)
/* preserve volumes -- always */
| preserve_reinstall:([[:digit:]]+(,[[:digit:]]+)*|all)
/* preserve volumes -- unless the system is installed for the
first time */
| preserve_lazy:([[:digit:]]+(,[[:digit:]]+)*|all)
/* preserve volumes -- unless these don’t exist yet */
| always_format:([[:digit:]]+(,[[:digit:]]+)*|all)
/* run mkfs on the volumes, even if marked as preserve */
| fstabkey:(device|label|uuid)
/* when creating the fstab the key used for defining the device
may be the device (/dev/xxx), a label given using -L, or the uuid

cryptsetupoption ::= /* empty */
| randinit
/* initialise all encrypted partitions with random data */

option ::= /* empty */
| preserve_always:([[:digit:]]+(,[[:digit:]]+)*|all)
/* preserve partitions -- always; the numbers refer to partition
numbers, i.e., preserve_always:5 for /dev/hda refers to /dev/hda5,
which may not necessarily be the 5th line of the configuration */
| preserve_reinstall:([[:digit:]]+(,[[:digit:]]+)*|all)
/* preserve partitions -- unless the system is installed for the
first time. See preserve_always above for the semantics of numbers
used for referring to partitions. */
| preserve_lazy:([[:digit:]]+(,[[:digit:]]+)*|all)
/* preserve partitions -- unless these don’t exist yet */
| always_format:([[:digit:]]+(,[[:digit:]]+)*|all)
/* run mkfs on the partitions, even if marked as preserve */
| resize:([[:digit:]]+(,[[:digit:]]+)*|all)
/* attempt to resize partitions */
| disklabel:(msdos|gpt|gpt-bios)
/* write a disklabel - default is msdos */
| bootable:[[:digit:]]+
/* mark a partition bootable, default is / */
| virtual
/* do not assume the disk to be a physical device, use with xen */
| fstabkey:(device|label|uuid)
/* when creating the fstab the key used for defining the device
may be the device (/dev/xxx), a label given using -L, or the uuid
| sameas:(disk[[:digit:]]+|[^[:space:]]+)
/* Indicate that this disk will use the same scheme
as the given device. The referenced device must be
defined before the device using this option. Use only
with identical hardware.
| align-at:([[:digit:]]+[kKMGTPiB]*)
/* Align partitions at multiples of the given block size (unit
defaults to MiB, if omitted). Such an alignment, e.g., 4K, might be
important for proper performance of RAID arrays which use a logical
block size other than the sector size of the underlying disks. It
must, however, always be a multiple of this sector size.

volume ::= <type> <mountpoint> <size> <filesystem> <mount_options> <luks_options> <fs_options>
| vg <name> <size> <fs_options>
/* lvm vg */
| tmpfs <mountpoint> <tmpfs_size> <mount_options>
/* tmpfs volume */

type ::= primary
/* for physical disks only */
| logical
/* for physical disks only */
| raw-disk
/* for physical disks only: do not partition this disk, use it as-is */
| raid[0156]
/* raid level */
| luks
/* encrypted partition using LUKS and auto-generate a key file */
| luks:"[^"]+"
/* encrypted partition using LUKS and use quoted string as passphrase */
| tmp
/* encrypted partition for /tmp usage, will be
recreated with a random key at each boot and
reformatted as ext2 */
| swap
/* encrypted partition for swap space usage, will
be recreated with a random key at each boot and
reformatted as swap space */
| [^/[:space:]]+-[^/[:space:]]+
/* lvm logical volume: vg name and lv name*/

mountpoint ::= (-|swap|/[^[:space:]]*)
/* do not mount, mount as swap, or mount at fully qualified path */

name ::= [^/[:space:]]+
/* lvm volume group name */

sizespec ::= RAM:[[:digit:]]+%|[[:digit:]]+[kKMGTP%iB]*
/* size in kilo (KiB), mega (default, MiB), giga (GiB), tera (TiB),
* petabytes (PiB) or percentage of disk size or RAM size; integers
* only, no decimal numbers.
* Use KB, MB, GB, ... for a factor of 1000 instead of 1024 as
* multiplier */

size ::= <sizespec>(-(<sizespec>)?)?(:resize|:preserve_(always|reinstall|lazy))?
/* size, possibly given as a range; physical partitions or lvm logical
* volumes only */
| -<sizespec>(:resize|:preserve_(always|reinstall|lazy))?
/* size given as upper limit; physical partitions or lvm logical
* volumes only */
| [^,:[:space:]]+(:(spare|missing))*(,[^,:[:space:]]+(:(spare|missing))*)*
/* devices and options for a raid or lvm vg */

tmpfs_size ::= <sizespec>
/* tmpfs size */

mount_options ::= [^[:space:]]+

filesystem ::= -
| swap
| [^[:space:]]
/* must exist */

luks_options ::= (lukscreateopts=".*")
/* options to supply to cryptsetup when creating a LUKS
* encrypted filesystem. If no ciper (-c) is specified, then
* aes-xts-plain64 is used. If no key size (-s) is
* specified then 256 is used. */

fs_options ::= (createopts=".*"|tuneopts=".*"|(pv|vg|lv|md)createopts=".*")*
/* options to append to and to the filesystem-specific
* tuning tool, pvcreate, vgcreate, lvcreate or mdadm */


If you use software RAID or LVM in your disk config, do not forget also to add the required packages (like mdadm and lvm2) to the package config. Otherwise the client will not be able to use these features.

MD devices aka software RAID will not be synced during a new installation because it would slow down the installation heavily. Therefore setup-storage will set the sync_action of each device to frozen. The sync will start after the next reboot.

Machine does not boot because no partition is marked as bootable: If the bootable option is not specified, not partition will be marked as such. Modern BIOSes don’t seem to require such markers anymore, but for some systems it may still be necessary. Previous versions of setup-storage by default marked the partition mounting / as bootable, but this is not a sane default for all cases. If you want to be sure not boot failures happen because of a missing bootable marker, explicitly set the bootable option. Of course, there are lots of other reasons why a system may fail to boot.

Crypto support requires some site-specific changes: If you use cryptsetup stanza, a crypttab(5) file and key files for all luks volumes will be created (unless you used the passphrase option). The key files are left in /tmp/fai; you will want to copy these to some removable media. To make encrypted root devices actually usable, you need to add busybox (and initramfs-tools) to your package config.

For backwards compatibility or other system-specific reasons an alignment to cylinder boundaries may be necessary. Other systems might have other alignment constraints. setup-storage sets the alignment as follows: If align-at is set, align accordingly. Otherwise, if any partition on the particular disk is to be preserved, default to cylinder alignment, otherwise use sector alignment.

If you use a raw partition, you should add -F to the createopts, otherwise mkfs.ext4 hangs due to a warning message. See man mfs.ext4 for more info on -F. An example config using a raw disk looks like this: raw-disk / 4GB ext4 rw createopts="-F"


This program is part of FAI (Fully Automatic Installation). The FAI homepage is

Further documentation, including coding related information, is available in a wiki page at


The setup-storage program was written by Michael Tautschnig <>, with contributions from Christian Kern, Julien Blache <>, Kerim Güney, Thomas Lange and others.

The original and primary author of FAI is Thomas Lange <>.