lorax-composer

Authors

Brian C. Lane <bcl@redhat.com>

lorax-composer is a WELDR API server that allows you to build disk images using Blueprints to describe the package versions to be installed into the image. It is compatible with the Weldr project's bdcs-api REST protocol. More information on Weldr can be found on the Weldr blog.

Behind the scenes it uses livemedia-creator and Anaconda to handle the installation and configuration of the images.

Note

lorax-composer is now deprecated. It is being replaced by the osbuild-composer WELDR API server which implements more features (eg. ostree, image uploads, etc.) You can still use composer-cli and cockpit-composer with osbuild-composer. See the documentation or the osbuild website for more information.

Important Things To Note

  • As of version 30.7 SELinux can be set to Enforcing. The current state is logged for debugging purposes and if there are SELinux denials they should be reported as a bug.

  • All image types lock the root account, except for live-iso. You will need to either use one of the Customizations methods for setting a ssh key/password, install a package that creates a user, or use something like cloud-init to setup access at boot time.

Installation

The best way to install lorax-composer is to use sudo dnf install lorax-composer composer-cli, this will setup the weldr user and install the systemd socket activation service. You will then need to enable it with sudo systemctl enable lorax-composer.socket && sudo systemctl start lorax-composer.socket. This will leave the server off until the first request is made. Systemd will then launch the server and it will remain running until the system is rebooted. This will cause some delay in responding to the first request from the UI or composer-cli.

Note

If you want lorax-composer to respond immediately to the first request you can start and enable lorax-composer.service instead of lorax-composer.socket

Quickstart

  1. Create a weldr user and group by running useradd weldr

  2. Remove any pre-existing socket directory with rm -rf /run/weldr/ A new directory with correct permissions will be created the first time the server runs.

  3. Enable the socket activation with systemctl enable lorax-composer.socket && sudo systemctl start lorax-composer.socket.

NOTE: You can also run it directly with lorax-composer /path/to/blueprints. However, lorax-composer does not react well to being started both on the command line and via socket activation at the same time. It is therefore recommended that you run it directly on the command line only for testing or development purposes. For real use or development of other projects that simply use the API, you should stick to socket activation only.

The /path/to/blueprints/ directory is where the blueprints' git repo will be created, and all the blueprints created with the /api/v0/blueprints/new route will be stored. If there are blueprint .toml files in the top level of the directory they will be imported into the blueprint git storage when lorax-composer starts.

Logs

Logs are stored under /var/log/lorax-composer/ and include all console messages as well as extra debugging info and API requests.

Security

Some security related issues that you should be aware of before running lorax-composer:

  • One of the API server threads needs to retain root privileges in order to run Anaconda.

  • Only allow authorized users access to the weldr group and socket.

Since Anaconda kickstarts are used there is the possibility that a user could inject commands into a blueprint that would result in the kickstart executing arbitrary code on the host. Only authorized users should be allowed to build images using lorax-composer.

lorax-composer cmdline arguments

Lorax Composer API Server

usage: lorax-composer [-h] [--socket SOCKET] [--user USER] [--group GROUP] [--log LOG] [--mockfiles MOCKFILES] [--sharedir SHAREDIR] [-V] [-c CONFIG] [--releasever STRING] [--tmp TMP] [--proxy PROXY] [--no-system-repos] BLUEPRINTS

Positional Arguments

BLUEPRINTS

Path to the blueprints

Named Arguments

--socket

Path to the socket file to listen on

Default: "/run/weldr/api.socket"

--user

User to use for reduced permissions

Default: "root"

--group

Group to set ownership of the socket to

Default: "weldr"

--log

Path to logfile (/var/log/lorax-composer/composer.log)

Default: "/var/log/lorax-composer/composer.log"

--mockfiles

Path to JSON files used for /api/mock/ paths (/var/tmp/bdcs-mockfiles/)

Default: "/var/tmp/bdcs-mockfiles/"

--sharedir

Directory containing all the templates. Overrides config file sharedir

-V

show program's version number and exit

Default: False

-c, --config

Path to lorax-composer configuration file.

Default: "/etc/lorax/composer.conf"

--releasever

Release version to use for $releasever in dnf repository urls

--tmp

Top level temporary directory

Default: "/var/tmp"

--proxy

Set proxy for DNF, overrides configuration file setting.

--no-system-repos

Do not copy over system repos from /etc/yum.repos.d/ at startup

Default: False

How it Works

The server runs as root, and as weldr. Communication with it is via a unix domain socket (/run/weldr/api.socket by default). The directory and socket are owned by root:weldr so that any user in the weldr group can use the API to control lorax-composer.

At startup the server will check for the correct permissions and ownership of a pre-existing directory, or it will create a new one if it doesn't exist. The socket path and group owner's name can be changed from the cmdline by passing it the --socket and --group arguments.

It will then drop root privileges for the API thread and run as the weldr user. The queue and compose thread still runs as root because it needs to be able to mount/umount files and run Anaconda.

Composing Images

The welder-web GUI project can be used to construct blueprints and create composes using a web browser.

Or use the command line with composer-cli.

Blueprints

Blueprints are simple text files in TOML format that describe which packages, and what versions, to install into the image. They can also define a limited set of customizations to make to the final image.

Example blueprints can be found in the lorax-composer test suite, with a simple one looking like this:

name = "base"
description = "A base system with bash"
version = "0.0.1"

[[packages]]
name = "bash"
version = "4.4.*"

The name field is the name of the blueprint. It can contain spaces, but they will be converted to - when it is written to disk. It should be short and descriptive.

description can be a longer description of the blueprint, it is only used for display purposes.

version is a semver compatible version number. If a new blueprint is uploaded with the same version the server will automatically bump the PATCH level of the version. If the version doesn't match it will be used as is. eg. Uploading a blueprint with version set to 0.1.0 when the existing blueprint version is 0.0.1 will result in the new blueprint being stored as version 0.1.0.

[[packages]] and [[modules]]

These entries describe the package names and matching version glob to be installed into the image.

The names must match the names exactly, and the versions can be an exact match or a filesystem-like glob of the version using * wildcards and ? character matching.

NOTE: Currently there are no differences between packages and modules in lorax-composer. Both are treated like an rpm package dependency.

For example, to install tmux-2.9a and openssh-server-8.*, you would add this to your blueprint:

[[packages]]
name = "tmux"
version = "2.9a"

[[packages]]
name = "openssh-server"
version = "8.*"

[[groups]]

The groups entries describe a group of packages to be installed into the image. Package groups are defined in the repository metadata. Each group has a descriptive name used primarily for display in user interfaces and an ID more commonly used in kickstart files. Here, the ID is the expected way of listing a group.

Groups have three different ways of categorizing their packages: mandatory, default, and optional. For purposes of blueprints, mandatory and default packages will be installed. There is no mechanism for selecting optional packages.

For example, if you want to install the anaconda-tools group you would add this to your blueprint:

[[groups]]
name="anaconda-tools"

groups is a TOML list, so each group needs to be listed separately, like packages but with no version number.

Customizations

The [customizations] section can be used to configure the hostname of the final image. eg.:

[customizations]
hostname = "baseimage"

This is optional and may be left out to use the defaults.

[customizations.kernel]

This allows you to append arguments to the bootloader's kernel commandline. This will not have any effect on tar or ext4-filesystem images since they do not include a bootloader.

For example:

[customizations.kernel]
append = "nosmt=force"

[[customizations.sshkey]]

Set an existing user's ssh key in the final image:

[[customizations.sshkey]]
user = "root"
key = "PUBLIC SSH KEY"

The key will be added to the user's authorized_keys file.

Warning

key expects the entire content of ~/.ssh/id_rsa.pub

[[customizations.user]]

Add a user to the image, and/or set their ssh key. All fields for this section are optional except for the name, here is a complete example:

[[customizations.user]]
name = "admin"
description = "Administrator account"
password = "$6$CHO2$3rN8eviE2t50lmVyBYihTgVRHcaecmeCk31L..."
key = "PUBLIC SSH KEY"
home = "/srv/widget/"
shell = "/usr/bin/bash"
groups = ["widget", "users", "wheel"]
uid = 1200
gid = 1200

If the password starts with $6$, $5$, or $2b$ it will be stored as an encrypted password. Otherwise it will be treated as a plain text password.

Warning

key expects the entire content of ~/.ssh/id_rsa.pub

[[customizations.group]]

Add a group to the image. name is required and gid is optional:

[[customizations.group]]
name = "widget"
gid = 1130

[customizations.timezone]

Customizing the timezone and the NTP servers to use for the system:

[customizations.timezone]
timezone = "US/Eastern"
ntpservers = ["0.north-america.pool.ntp.org", "1.north-america.pool.ntp.org"]

The values supported by timezone can be listed by running timedatectl list-timezones.

If no timezone is setup the system will default to using UTC. The ntp servers are also optional and will default to using the distribution defaults which are fine for most uses.

In some image types there are already NTP servers setup, eg. Google cloud image, and they cannot be overridden because they are required to boot in the selected environment. But the timezone will be updated to the one selected in the blueprint.

[customizations.locale]

Customize the locale settings for the system:

[customizations.locale]
languages = ["en_US.UTF-8"]
keyboard = "us"

The values supported by languages can be listed by running localectl list-locales from the command line.

The values supported by keyboard can be listed by running localectl list-keymaps from the command line.

Multiple languages can be added. The first one becomes the primary, and the others are added as secondary. One or the other of languages or keyboard must be included (or both) in the section.

[customizations.firewall]

By default the firewall blocks all access except for services that enable their ports explicitly, like sshd. This command can be used to open other ports or services. Ports are configured using the port:protocol format:

[customizations.firewall]
ports = ["22:tcp", "80:tcp", "imap:tcp", "53:tcp", "53:udp"]

Numeric ports, or their names from /etc/services can be used in the ports enabled/disabled lists.

The blueprint settings extend any existing settings in the image templates, so if sshd is already enabled it will extend the list of ports with the ones listed by the blueprint.

If the distribution uses firewalld you can specify services listed by firewall-cmd --get-services in a customizations.firewall.services section:

[customizations.firewall.services]
enabled = ["ftp", "ntp", "dhcp"]
disabled = ["telnet"]

Remember that the firewall.services are different from the names in /etc/services.

Both are optional, if they are not used leave them out or set them to an empty list []. If you only want the default firewall setup this section can be omitted from the blueprint.

NOTE: The Google and OpenStack templates explicitly disable the firewall for their environment. This cannot be overridden by the blueprint.

[customizations.services]

This section can be used to control which services are enabled at boot time. Some image types already have services enabled or disabled in order for the image to work correctly, and cannot be overridden. eg. ami requires sshd, chronyd, and cloud-init. Without them the image will not boot. Blueprint services are added to, not replacing, the list already in the templates, if any.

The service names are systemd service units. You may specify any systemd unit file accepted by systemctl enable eg. cockpit.socket:

[customizations.services]
enabled = ["sshd", "cockpit.socket", "httpd"]
disabled = ["postfix", "telnetd"]

[[repos.git]]

The [[repos.git]] entries are used to add files from a git repository repository to the created image. The repository is cloned, the specified ref is checked out and an rpm is created to install the files to a destination path. The rpm includes a summary with the details of the repository and reference used to create it. The rpm is also included in the image build metadata.

To create an rpm named server-config-1.0-1.noarch.rpm you would add this to your blueprint:

[[repos.git]]
rpmname="server-config"
rpmversion="1.0"
rpmrelease="1"
summary="Setup files for server deployment"
repo="PATH OF GIT REPO TO CLONE"
ref="v1.0"
destination="/opt/server/"
  • rpmname: Name of the rpm to create, also used as the prefix name in the tar archive

  • rpmversion: Version of the rpm, eg. "1.0.0"

  • rpmrelease: Release of the rpm, eg. "1"

  • summary: Summary string for the rpm

  • repo: URL of the get repo to clone and create the archive from

  • ref: Git reference to check out. eg. origin/branch-name, git tag, or git commit hash

  • destination: Path to install the / of the git repo at when installing the rpm

An rpm will be created with the contents of the git repository referenced, with the files being installed under /opt/server/ in this case.

ref can be any valid git reference for use with git archive. eg. to use the head of a branch set it to origin/branch-name, a tag name, or a commit hash.

Note that the repository is cloned in full each time a build is started, so pointing to a repository with a large amount of history may take a while to clone and use a significant amount of disk space. The clone is temporary and is removed once the rpm is created.

Adding Output Types

livemedia-creator supports a large number of output types, and only some of these are currently available via lorax-composer. To add a new output type to lorax-composer a kickstart file needs to be added to ./share/composer/. The name of the kickstart is what will be used by the /compose/types route, and the compose_type field of the POST to start a compose. It also needs to have code added to the pylorax.api.compose.compose_args() function. The _MAP entry in this function defines what lorax-composer will pass to pylorax.installer.novirt_install() when it runs the compose. When the compose is finished the output files need to be copied out of the build directory (/var/lib/lorax/composer/results/<UUID>/compose/), pylorax.api.compose.move_compose_results() handles this for each type. You should move them instead of copying to save space.

If the new output type does not have support in livemedia-creator it should be added there first. This will make the output available to the widest number of users.

Example: Add partitioned disk support

Partitioned disk support is something that livemedia-creator already supports via the --make-disk cmdline argument. To add this to lorax-composer it needs 3 things:

The partitioned-disk.ks is pretty similar to the example minimal kickstart in ./docs/fedora-minimal.ks. You should remove the url and repo commands, they will be added by the compose process. Make sure the bootloader packages are included in the %packages section at the end of the kickstart, and you will want to leave off the %end so that the compose can append the list of packages from the blueprint.

The new _MAP entry should be a copy of one of the existing entries, but with make_disk set to True. Make sure that none of the other make_* options are True. The image_name is what the name of the final image will be.

move_compose_results() can be as simple as moving the output file into the results directory, or it could do some post-processing on it. The end of the function should always clean up the ./compose/ directory, removing any unneeded extra files. This is especially true for the live-iso since it produces the contents of the iso as well as the boot.iso itself.

Package Sources

By default lorax-composer uses the host's configured repositories. It copies the *.repo files from /etc/yum.repos.d/ into /var/lib/lorax/composer/repos.d/ at startup, these are immutable system repositories and cannot be deleted or changed. If you want to add additional repos you can put them into /var/lib/lorax/composer/repos.d/ or use the /api/v0/projects/source/* API routes to create them.

The new source can be added by doing a POST to the /api/v0/projects/source/new route using JSON (with Content-Type header set to application/json) or TOML (with it set to text/x-toml). The format of the source looks like this (in TOML):

name = "custom-source-1"
url = "https://url/path/to/repository/"
type = "yum-baseurl"
proxy = "https://proxy-url/"
check_ssl = true
check_gpg = true
gpgkey_urls = ["https://url/path/to/gpg-key"]

The proxy and gpgkey_urls entries are optional. All of the others are required. The supported types for the urls are:

  • yum-baseurl is a URL to a yum repository.

  • yum-mirrorlist is a URL for a mirrorlist.

  • yum-metalink is a URL for a metalink.

If check_ssl is true the https certificates must be valid. If they are self-signed you can either set this to false, or add your Certificate Authority to the host system.

If check_gpg is true the GPG key must either be installed on the host system, or gpgkey_urls should point to it.

You can edit an existing source (other than system sources), by doing a POST to the new route with the new version of the source. It will overwrite the previous one.

A list of existing sources is available from /api/v0/projects/source/list, and detailed info on a source can be retrieved with the /api/v0/projects/source/info/<source-name> route. By default it returns JSON but it can also return TOML if ?format=toml is added to the request.

Non-system sources can be deleted by doing a DELETE request to the /api/v0/projects/source/delete/<source-name> route.

The documentation for the source API routes can be found here

The configured sources are used for all blueprint depsolve operations, and for composing images. When adding additional sources you must make sure that the packages in the source do not conflict with any other package sources, otherwise depsolving will fail.

DVD ISO Package Source

In some situations the system may want to only use a DVD iso as the package source, not the repos from the network. lorax-composer and anaconda understand file:// URLs so you can mount an iso on the host, and replace the system repo files with a configuration file pointing to the DVD.

  • Stop the lorax-composer.service if it is running

  • Move the repo files in /etc/yum.repos.d/ someplace safe

  • Create a new iso.repo file in /etc/yum.repos.d/:

    [iso]
    name=iso
    baseurl=file:///mnt/iso/
    enabled=1
    gpgcheck=1
    gpgkey=file:///mnt/iso/RPM-GPG-KEY-redhat-release
    
  • Remove all the cached repo files from /var/lib/lorax/composer/repos/

  • Restart the lorax-composer.service

  • Check the output of composer-cli status show for any output specific depsolve errors. For example, the DVD usually does not include grub2-efi-*-cdboot-* so the live-iso image type will not be available.

If you want to add the DVD source to the existing sources you can do that by mounting the iso and creating a source file to point to it as described in the Package Sources documentation. In that case there is no need to remove the other sources from /etc/yum.repos.d/ or clear the cached repos.