README for vmdb2 or vmdebootstrap 2nd generation ; -*- mode: markdown;-*- ============================================================================= **Note:** vmdb2 is in "selfish maintenance mode". Lars maintains the software to the extent he needs it, but is not spending time to develop new features or debug problems he doesn't see himself. He will review patches, however, so if you want vmdb2 to improve, make a change and submit it for review. [vmdb2][] is a program for producing a disk image with Debian installed. [vmdb2]: https://vmdb2.liw.fi/ Introduction ----------------------------------------------------------------------------- [vmdebootstrap][] installs Debian onto a disk image. It is like the [debootstrap][] tool, except the end result is a bootable disk image, not a directory. vmdebootstrap takes care of creating partitions, and filesystems, and allows some more customization than the older vmdebootstrap does. vmdebootstrap is also a messy pile of kludge, and rather inflexible. vmdb2 is a re-implementation from scratch, without a need for backwards compatibility. It aims to provide more flexibility than vmdebootstrap, without becoming anywhere near as complicated. Think of vmdb2 as "vmdebootstrap the second generation". The name has changed to allow the two tools to installable in parallel, which is important for a transition period. The main user-visible difference between vmdebootstrap and vmdb2 is that the older program provides extensibility via a legion of command line options and the newer program by providing a domain specific language to express what kind of Debian system is to be created. (Lars Wirzenius wrote both vmdebootstrap and vmdb2 and is entitled to sneer at his younger self. It's his way of dealing with the mountain of guilt of making something as awful as vmdebootstrap.) [vmdebootstrap]: http://liw.fi/vmdebootstrap/ [debootstrap]: https://packages.debian.org/unstable/debootstrap Getting vmdb2 ----------------------------------------------------------------------------- vmdb2 source code is available via git: * * It used to be on GitHub as well, but was withdrawn from there due to GitHub being a proprietary service. Requirements: The following tools are used by vmdb2 (Debian package names in brackets). * `kpartx` [kpartx, mkpart command] * `parted` [`parted`, mklabel command] * `qemu-img` [`qemu-utils`, mkimg command] * `qemu-user-static` [`qemu-user-static`] * `zerofree` [`zerofree`, zerofree command] The following Python modules are used by vmdb2 (Debian package names in brackets). * jinja2 [`python3-jinja2`] * yaml [`python3-yaml`] If UEFI booting is to be used, firmware is needed for the following architectures (Debian package names providing it in brackets): * amd64 [`ovmf`] * arm64 [`qemu-efi-aarch64`] * arm [`qemu-efi-arm`] * i386 [`ovmf-ia32`] Dependencies for smoke.sh ----------------------------------------------------------------------------- You probably need the following installed to run the smoke tests: - git - python3-coverage-test-runner - python3-jinja2 - cmdtest 0.31 or later - qemu-utils - parted - kpartx - debootstrap - expect - qemu-system - ovmf - ovmf-ia32 - qemu-efi-aarch64 - qemu-efi-arm - zerofree Tutorial ----------------------------------------------------------------------------- To use vmdb2, git clone the source and at the root of the source tree run the following command: sudo ./vmdb2 --output test.img base.vmdb --log vmdb2.log `--output test.img` specifies that the output image is called `test.img`, the specification is `base.vmdb` and the log file goes to `vmdb2.log`. Plugins and steps ----------------------------------------------------------------------------- The `vmdb2` architecture consists of a main program that reads the input file, finds a matching "step runner" for each step used in the input file, and then runs the steps in order. If there's a problem, it runs corresponding "teardown" steps in reverse order of the steps. A step might be "mount this filesystem", and the corresponding teardown is "unmount". Steps (and teardowns) are provided by plugins; see the `vmdb/plugins` directory in the source tree. Steps are intended to be very cohesive and lowly coupled. They may share some state (such as mounted filesystems) via the `State` object, but not in any other way. A plugin may only provide one step runner. See `base.vmdb` and other `.vmdb` files for examples. Note how the file uses Jinja2 templating for value fields to get value of `--output` in the right places. Also note how creating a partition or mounting a filesystem assigns a "tag" that can be referenced in steps where the partition/filesystem is needed, without having to know the actual path to the device node or mount point. Writing plugins ----------------------------------------------------------------------------- More step runners would be good, but will be added based on actual reported needs by users ("I need to have this to..."), not speculatively ("This seems like a good idea"). To write a plugin, see the existing ones for examples, and put it in `vmdb/plugins/foo_plugin.py` for some value of `foo`. The plugin file should provide a class named `FooPlugin`, which should provide the interface defined by `vmdb.Plugin`. Most plugins add a step runner, which subclasses `vmdb.StepRunnerInterface`, and the plugin class adds the step runner to the application's global list of step runners. See existing plugins for examples. Note that each plugin may only add one step runner. This keeps things simple, and also keeps document formatting simple. If two plugins need to share code for some reason, it may be appropriate to put that code into the `vmdb` module. You should document the plugin in a Markdown file next to it: `vmdb/plugins/foo.mdwn` for plugin mentioned above. See existing documentation files for a model. You should mention all keys each step can use, and give an example. It would be great to explain when the plugin would be useful. Try to keep source lines to less than 80 characters. Plugins are meant to be very easy to write. If not, there's probably something wrong with `vmdb2`. Please raise the issue. Hacking ----------------------------------------------------------------------------- To run automated tests: ./check This only runs the unit tests and build tests. To run a smoke test that actually builds and boots images: sudo ./smoke.sh cache.tar.gz where `cache.tar.gz` caches the debootstrap output for a future run. You'll need the yarn program (part of the [cmdtest][] package), and also [CoverageTestRunner][] for running the unit tests. [cmdtest]: http://liw.fi/cmdtest/ [CoverageTestRunner]: http://liw.fi/coverage-test-runner/ Try to follow PEP8 for code formatting, and try to keep lines shorter than 80 characters. Make sure `./check-all` script passes both before and after your modifications. Contact ----------------------------------------------------------------------------- To contact Lars, email is best: `liw@liw.fi`. There is an IRC channel for vmdb2: irc.oftc.net network, `#vmdb2`. Legalese ----------------------------------------------------------------------------- Copyright 2017-2019 Lars Wirzenius This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version. This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details. You should have received a copy of the GNU General Public License along with this program. If not, see . =*= License: GPL-3+ =*=