Table of Contents
- About
- History
- Example run
- Speeding things up
- Debian package
- Usage (
--helpoutput) - Hints on using image-bootstrap within a pipe
- Known limitations
Welcome to the home of image-bootstrap (and its little brother directory-bootstrap).
image-bootstrap is a command line tool to generate bootable virtual machine images
and write them to a given block device.
Linux distributions supported by image-bootstrap currently include:
Arch, Debian, Gentoo, Ubuntu.
When passing the --openstack parameter, images are
prepared for use with OpenStack.
directory-bootstrap is a command line tool to install non-Debian Linux distributions
into a given directory (similar to debootstrap
for Debian/Ubuntu).
Distributions supported by directory-bootstrap currently include:
Alpine Linux, Arch Linux, Gentoo, and Void Linux.
image-bootstrap started out as a re-write of grml-debootstrap. Primarily, it can be used to create Debian/Ubuntu or Arch images ready to be launched as a virtual machine.
In comparison to grml-debootstrap, by now image-bootstrap
-
installs to block devices only,
-
supports several approaches to installing GRUB 2.x, or extlinux, or no bootloader at all,
-
does not leak the host's hostname into the resulting image,
-
supports passing the root password off the command-line (and the eyes of other users and shell history),
-
supports using a custom
/etc/resolv.conffile (e.g. to not leak your home router model name from/etc/resolv.confinto the image), -
has support for installing Arch Linux, Gentoo and Ubuntu (besides Debian),
-
is able to create OpenStack images,
-
is written in Python rather than Bash/mksh, and
-
has more friendly terminal output.
directory-bootstrap came into life with the arrival of support for Arch Linux.
Support for Gentoo followed, after.
Support for Void Linux and CentOS chroots came into live during 34c3, December 2017.
Support for Alpine Linux chroots came into in March 2018.
The following is a complete demo of installing Debian stretch to LVM volume /dev/vg/lv
and launching the resulting image using KVM.
# ${EDITOR} root_password.txt
# sudo image-bootstrap --hostname stretch debian \
--password-file root_password.txt /dev/vg/lv
_ __ __ __
(_)_ _ ___ ____ ____ ___ / / ___ ___ / /____ / /________ ____
/ / ' \/ _ `/ _ `/ -_)/__// _ \/ _ \/ _ \/ __(_-</ __/ __/ _ `/ _ \
/_/_/_/_/\_,_/\_, /\__/ /_.__/\___/\___/\__/___/\__/_/ \_,_/ .__/
/___/ v0.9.1 :: 2015-07-11 /_/
Software libre licensed under AGPL v3 or later.
Brought to you by Sebastian Pipping <sebastian@pipping.org>.
Please report bugs at https://github.com/hartwork/image-bootstrap. Thank you!
Selected approach "chroot-grub2-drive" for bootloader installation.
Checking for blkid... /sbin/blkid
Checking for chmod... /bin/chmod
Checking for chroot... /usr/sbin/chroot
Checking for cp... /bin/cp
Checking for debootstrap... /usr/sbin/debootstrap
Checking for find... /usr/bin/find
Checking for kpartx... /sbin/kpartx
Checking for mkdir... /bin/mkdir
Checking for mkfs.ext4... /sbin/mkfs.ext4
Checking for mount... /bin/mount
Checking for parted... /sbin/parted
Checking for partprobe... /sbin/partprobe
Checking for rm... /bin/rm
Checking for rmdir... /bin/rmdir
Checking for sed... /bin/sed
Checking for tune2fs... /sbin/tune2fs
Checking for umount... /bin/umount
Checking for uname... /bin/uname
Checking for unshare... /usr/bin/unshare
Checking for known unsupported architecture/machine combination...
Checking if "/dev/vg/lv" is a block device...
Reading root password from file "/home/user1/root_password.txt"...
Unsharing Linux namespaces (mount, UTS/hostname)...
Partitioning "/dev/vg/lv"...
Activating partition devices...
Creating file system on "/dev/mapper/vg-lvp1"...
Creating directory "/mnt/tmpFczeFl"...
Mounting partitions...
Creating directory "/mnt/tmpFczeFl/etc"...
Writing file "/mnt/tmpFczeFl/etc/hostname"...
Writing file "/mnt/tmpFczeFl/etc/resolv.conf" (based on file "/etc/resolv.conf")...
Bootstrapping Debian "stretch" into "/mnt/tmpFczeFl"...
Writing file "/mnt/tmpFczeFl/etc/hostname"...
Writing file "/mnt/tmpFczeFl/etc/resolv.conf" (based on file "/etc/resolv.conf")...
Writing file "/mnt/tmpFczeFl/etc/fstab"...
Writing file "/mnt/tmpFczeFl/etc/network/interfaces"...
Running pre-chroot scripts...
Mounting non-disk file systems...
Setting root password...
Writing device map to "/mnt/tmpFczeFl/boot/grub/device.map" (mapping "(hd9999)" to "/dev/dm-8")...
Installing bootloader to device "/dev/vg/lv" (actually "/dev/dm-8", approach "chroot-grub2-drive")...
Generating GRUB configuration...
Post-processing GRUB config...
Generating initramfs...
Unmounting non-disk file systems...
Cleaning chroot apt cache...
Running post-chroot scripts...
Unmounting partitions...
Removing directory "/mnt/tmpFczeFl"...
Deactivating partition devices...
Done.
# sudo kvm -hda /dev/vg/lvWithout --color never, the output above is actually in color.
If you run image-bootstrap repeatedly and have enough RAM, you may want to create images on RAM storage rather than on disk. I use a setup with
-
a loop device (to have a block device)
-
over a sparse file (to save space)
-
in a tmpfs mount (to use RAM).
For example (assuming you have /tmp in RAM already):
# sudo mount -o remount,size=6g /tmp
# truncate --size 3g /tmp/disk3g
# LOOP_DEVICE="$(losetup --show -f /tmp/disk3g)"
# image-bootstrap .... arch ... "${LOOP_DEVICE}"
# qemu-img convert -p -f raw -O qcow2 "${LOOP_DEVICE}" /var/lib/arch-$(date -I).qcow2
# losetup -d "${LOOP_DEVICE}"
# rm /tmp/disk3gWhen creating multiple images,
a local instance of Apt-Cacher NG and
passing --mirror http://localhost:3142/debian to image-bootstrap may come in handy.
For a distribution-agnostic cache, using Polipo can greatly speed up consecutive runs. Invoke image-bootstrap with
# http_proxy=http://127.0.0.1:8123/ image-bootstrap ...when using Polipo with default port configuration.
During image creation, cryptographic keys may need to be generated, e.g. for the OpenSSH server, at least temporarily. As key generation relies on availability of entropy, image creation may take longer in environments that are slow at adding to the entropy pool. To speed things up, running haveged at the host system could be an option, especially since all keys should be deleted from images, eventually. Otherwise, there is a risk of ending up with multiple systems having the same key allowing for attacks. I am unsure of the quality of entropy that haveged produces. Use is at your own risk.
As long as image-bootstrap as not available in Debian, you can make an image-bootstrap Debian package yourself easily from Git as follows:
# git clone https://github.com/hartwork/image-bootstrap.git
Cloning into 'image-bootstrap'...
[..]
# make -C image-bootstrap/ deb
[..]
# ls *.deb
image-bootstrap_0.9.1_all.deb
# sudo dpkg -i image-bootstrap_0.9.1_all.deb
[..]In general, the usage is:
image-bootstrap [..] DISTRIBUTION [..] DEVICEor
image-bootstrap --hostname NAME [DISTRO_AGNOSTIC] DISTRIBUTION [DISTRO_SPECIFIC] DEVICEin a bit more detail.
A dump of the current --help output would be:
# image-bootstrap --help
usage: image-bootstrap [-h] [--version] [--color {never,always,auto}]
[--debug] [--quiet] [--verbose] [--arch ARCHITECTURE]
[--bootloader {auto,chroot-grub2-device,chroot-grub2-drive,host-extlinux,host-grub2-device,host-grub2-drive,none}]
[--bootloader-force] [--hostname NAME] [--openstack]
[--password PASSWORD | --password-file FILE]
[--resolv-conf FILE] [--disk-id ID]
[--first-partition-uuid UUID] [--machine-id ID]
[--scripts-pre DIRECTORY] [--scripts-chroot DIRECTORY]
[--scripts-post DIRECTORY] [--grub2-install COMMAND]
[--cache-dir DIRECTORY]
DISTRIBUTION ... DEVICE
Command line tool for creating bootable virtual machine images
positional arguments:
DEVICE block device to install to
optional arguments:
-h, --help show this help message and exit
--version show program's version number and exit
text output configuration:
--color {never,always,auto}
toggle output color (default: auto)
--debug enable debugging
--quiet limit output to error messages
--verbose increase verbosity
machine configuration:
--arch ARCHITECTURE architecture (e.g. amd64)
--bootloader {auto,chroot-grub2-device,chroot-grub2-drive,host-extlinux,host-grub2-device,host-grub2-drive,none}
approach to take during bootloader installation
(default: auto)
--bootloader-force apply more force when installing bootloader (default:
disabled)
--hostname NAME hostname to set (default: "machine")
--openstack prepare for use with OpenStack (default: disabled)
--password PASSWORD root password to set (default: password log-in
disabled)
--password-file FILE file to read root password from (default: password
log-in disabled)
--resolv-conf FILE file to copy nameserver entries from (default:
/etc/resolv.conf)
--disk-id ID specific disk identifier to apply, e.g. 0x12345678
--first-partition-uuid UUID
specific UUID to apply to first partition, e.g.
c1b9d5a2-f162-11cf-9ece-0020afc76f16
--machine-id ID specific machine identifier to apply, e.g.
c1b9d5a2f16211cf9ece0020afc76f16
script integration:
--scripts-pre DIRECTORY
scripts to run prior to chrooting phase, in
alphabetical order
--scripts-chroot DIRECTORY
scripts to run during chrooting phase, in alphabetical
order
--scripts-post DIRECTORY
scripts to run after chrooting phase, in alphabetical
order
command names:
--grub2-install COMMAND
override grub2-install command
general configuration:
--cache-dir DIRECTORY
directory to use for downloads (default:
/var/cache/directory-bootstrap/)
subcommands (choice of distribution):
Run "image-bootstrap DISTRIBUTION --help" for details on options specific to that distribution.
DISTRIBUTION choice of distribution, pick from:
arch Arch Linux
debian Debian GNU/Linux
gentoo Gentoo
ubuntu Ubuntu
_ __ __ __
(_)_ _ ___ ____ ____ ___ / / ___ ___ / /____ / /________ ____
/ / ' \/ _ `/ _ `/ -_)/__// _ \/ _ \/ _ \/ __(_-</ __/ __/ _ `/ _ \
/_/_/_/_/\_,_/\_, /\__/ /_.__/\___/\___/\__/___/\__/_/ \_,_/ .__/
/___/ v2.0.0 :: 2020-02-28 /_/
Software libre licensed under AGPL v3 or later.
Brought to you by Sebastian Pipping <sebastian@pipping.org>.
Please report bugs at https://github.com/hartwork/image-bootstrap. Thank you!To show options specific to Debian, run ..
# image-bootstrap debian --help
usage: image-bootstrap debian [-h] [--debootstrap COMMAND] [--release RELEASE]
[--mirror URL] [--debootstrap-opt OPTION]
optional arguments:
-h, --help show this help message and exit
--release RELEASE specify Debian release (default: stretch)
--mirror URL specify Debian mirror to use (e.g.
http://localhost:3142/debian for a local instance of
apt-cacher-ng; default:
http://httpredir.debian.org/debian)
--debootstrap-opt OPTION
option to pass to debootstrap, in addition; can be
passed several times; use with --debootstrap-opt=...
syntax, i.e. with "="
command names:
--debootstrap COMMAND
override debootstrap commandIf you want to run image-bootstrap in a pipe to capture its output to both
stdout/stderr to a single log file, be sure to run image-bootstrap in
unbuffered mode, e.g.:
python3 -u image-bootstrap [OPTIONS] 2>&1 | tee my.logThe default shebang generated by python setuptools does not use -u. It's also
not easily possible to have it pass -u since the env command does allow for
for passing command parameters only in fairly recent versions. As a consequence,
stdout/stderr won't be synchronized and error output you see in a log file
will not exactly correspond to preceding/succeeding output on stdout. (See
issue #71 for more details.)
Moreover, if you're using Bash and you need to keep track of
image-bootstrap's exit code, be sure to run set -o pipefail prior to
invoking image-bootstrap.
(Please see the
Pipelines
section of the
GNU Bash Reference Manual
for more details.)
Linux does not like partitions in partitions much. It can be tricked using device mapper, though.
This is how to install to a partition using another partition as a temporary target. The temporary target must
-
be 2 GiB in space or more (to hold the whole distribution) and
-
smaller or equal than the actually target (for the later copy to work).
# dmsetup create dm-linear-vda4 --table "0 $(blockdev --getsz /dev/vda4) linear /dev/vda4 0"
# image-bootstrap --openstack arch /dev/mapper/dm-linear-vda4
# partprobe /dev/mapper/dm-linear-vda4
# pv /dev/mapper/dm-linear-vda4p1 > /dev/vda2
# dmsetup remove dm-linear-vda4p1
# dmsetup remove dm-linear-vda4(/dev/vda2 is the real target, /dev/vda4 the temporary one.)
There are other ways to achieve the same.