Bifrost Installation


This document will guide you through installing the Bare Metal Service (ironic) using Bifrost.

Supported operating systems

1st tier support (fully tested in the CI, no known or potential issues):

  • CentOS Stream 8

  • Ubuntu 20.04 “Focal”

  • Debian 10 “Buster”

2nd tier support (limited testing or known issues):

  • Ubuntu 18.04 “Bionic”

    Tested in the Bifrost CI, but no longer tested in the ironic upstream CI.

  • RHEL 8 and regular CentOS 8

    Only tested indirectly via CentOS Stream 8.

  • openSUSE Leap 15.2

    Tested in the CI but has frequent issues.

  • Fedora 32 (30 is supported but not recommended)

    Only the latest Fedora is tested in the CI.


Operating systems evolve and so does the support for them, even on stable branches. This especially concerns Fedora, which is evolving faster than other distributions.

Bifrost structure

Installation and use of Bifrost is split into roughly three steps:

  • install: prepare the local environment by downloading and/or building machine images, and installing and configuring the necessary services.

  • enroll-dynamic: take as input a customizable hardware inventory file and enroll the listed hardware with ironic, configuring each appropriately for deployment with the previously-downloaded images.

  • deploy-dynamic: instruct ironic to deploy the operating system onto each machine.

Installation of Bifrost can be done in three ways:

  • Via the bifrost-cli command line tool.

    This is the path recommended for those who want something that just works. It provides minimum configuration and uses the recommended defaults.

  • By directly invoking ansible-playbook on one of provided playbooks.

  • By writing your own playbooks using Ansible roles provided with Bifrost.

Pre-install steps

Know your environment

Before you start, you need to gather certain facts about your bare metal environment (this step can be skipped if you’re testing Bifrost on virtual machines).

For the machine that hosts Bifrost you’ll need to figure out:

  • The network interface you’re going to use for communication between the bare metal machines and the Bifrost services.

    On systems using firewalld (Fedora, CentOS and RHEL currently), a new zone bifrost will be created, and the network interface will be moved to it. DHCP, PXE and API services will only be added to this zone. If you need any of them available in other zones, you need to configure firewall yourself.


    If you use the same NIC for bare metal nodes and external access, installing bifrost may lock you out of SSH to the node. You have two options:

    1. Pre-create the bifrost firewalld zone before installation and add the SSH service to it.

    2. Use the public zone by providing firewalld_internal_zone=public when installing.

  • Pool of IP addresses for DHCP (must be within the network configured on the chosen network interface).

  • Whether you want the services to use authentication via Keystone.

For each machine that is going to be enrolled in the Bare Metal service you’ll need:

  • The management technology you are going to use to control the machine (IPMI, Redfish, etc). See bare metal drivers for guidance.

  • An IP address or a host name of its management controller (BMC).

  • Credentials for the management controller.

  • MAC address of the NIC the machine uses for PXE booting (optional for IPMI).

Required packages

To start with Bifrost you will need Python 3.6 or newer and the git source code management tool.

On CentOS/RHEL/Fedora:

sudo dnf install -y git python3

On Ubuntu/Debian:

sudo apt-get update
sudo apt-get install -y python3 git

On openSUSE:

sudo zipper install -y python3 git

Enable additional repositories (RHEL only)

The extras and optional dnf repositories must be enabled to satisfy bifrost’s dependencies. To check:

sudo dnf repolist | grep 'optional\|extras'

To view the status of repositories:

sudo dnf repolist all | grep 'optional\|extras'

The output will look like this:

!rhui-REGION-rhel-server-debug-extras/8Server/x86_64        Red H disabled
rhui-REGION-rhel-server-debug-optional/8Server/x86_64       Red H disabled
rhui-REGION-rhel-server-extras/8Server/x86_64               Red H disabled
rhui-REGION-rhel-server-optional/8Server/x86_64             Red H disabled
rhui-REGION-rhel-server-source-extras/8Server/x86_64        Red H disabled
rhui-REGION-rhel-server-source-optional/8Server/x86_64      Red H disabled

Use the names of the repositories (minus the version and architecture) to enable them:

sudo dnf config-manager --enable rhui-REGION-rhel-server-optional
sudo dnf config-manager --enable rhui-REGION-rhel-server-extras

Enable the EPEL repository (RHEL and CentOS)

Building Debian or Ubuntu based images on RHEL or CentOS requires a few extra pre-install steps, in order to have access to the additional packages contained in the EPEL repository.

Please refer to the official wiki page to install and configure them.


Use of EPEL repositories may result in incompatible packages being installed by the package manager. Care should be taken when using a system with EPEL enabled.

Clone Bifrost

Bifrost is typically installed from git:

git clone
cd bifrost

To install Bare Metal services from a specific release series (rather than the latest versions), check out the corresponding stable branch. For example, for Ussuri:

git checkout stable/ussuri

Testing on virtual machines

If you want to try Bifrost on virtual machines instead of real hardware, you need to prepare a testing environment. The easiest way is via bifrost-cli, available since the Victoria release series:

./bifrost-cli testenv

Then do not forget to pass --testenv flag to bifrost-cli install.

See Testing Environment for more details and for advanced ways of creating a virtual environment (also supported on Ussuri and older).

Quick start with bifrost-cli

The bifrost-cli script, available since the Victoria release series, installs the Bare Metal service with the recommended defaults.


Follow Installation via playbooks if using Ussuri or older.

Using it is as simple as:

./bifrost-cli install \
    --network-interface <the network interface to use> \
    --dhcp-pool <DHCP start IP>-<DHCP end IP>

For example:

./bifrost-cli install --network-interface eno1 \


See Know your environment for the guidance on the two required parameters.

If installing on a virtual environment, skip these two parameters:

./bifrost-cli install --testenv

Additionally, the following parameters can be useful:


A comma separated list of hardware types to enable.


Whether to enable authentication with Keystone.


Enable self-signed TLS on API endpoints.


If using Keystone, see TLS notes for important notes.


If using a stable version of Bifrost, the corresponding version of Ironic is usually detected from the git checkout. If it is not possible (e.g. you’re using Bifrost from a tarball), use this argument to provide the matching version.


Using Bifrost to install older versions of Ironic may work, but is not guaranteed.


Enable the Ironic Prometheus Exporter service.

See the built-in documentation for more details:

./bifrost-cli install --help

Installation via playbooks

Installation is split into four parts:

  • Installation of Ansible

  • Configuring settings for the installation

  • Execution of the installation playbook

Installation of Ansible

Installation of Ansible can take place using the provided environment setup script located at scripts/ which is present in the bifrost repository. This may also be used if you already have ansible, as it will install ansible and various dependencies to a virtual environment in order to avoid overwriting or conflicting with a system-wide Ansible installation.

Alternatively, if you have a working Ansible installation, under normal circumstances the installation playbook can be executed, but you will need to configure the Virtual environment.


All testing takes place utilizing the scripts/ script. Please feel free to submit bug reports or patches to OpenStack Gerrit for any issues encountered if you choose to directly invoke the playbooks without using

Virtual environment

To avoid conflicts between Python packages installed from source and system packages, Bifrost defaults to installing everything to a virtual environment. scripts/ will automatically create a virtual environment in /opt/stack/bifrost if it does not exist.

If you want to relocate the virtual environment, export the VENV variable before calling

export VENV=/path/to/my/venv

If you’re using the ansible playbooks directly (without the helper scripts), set the bifrost_venv_dir variables accordingly.


Because of Ansible dependencies Bifrost only supports virtual environments created with --system-site-packages.

Pre-installation settings

Before performing the installation, it is highly recommended that you edit ./playbooks/inventory/group_vars/* to match your environment. Several files are located in this folder, and you may wish to review and edit the settings across multiple files:

  • The target file is used by roles that execute against the target node upon which you are installing ironic and all required services.

  • The baremetal file is geared for roles executed against baremetal nodes. This may be useful if you are automating multiple steps involving deployment and configuration of nodes beyond deployment via the same roles.

  • The localhost file is similar to the target file, and likely contains identical settings. This file is referenced if no explicit target is defined, as it defaults to the localhost.

Duplication between variable names does occur within these files, as variables are unique to the group that the role is being executed upon.

  • If MySQL is already installed, update mysql_password to match your local installation.

  • Change network_interface to match the interface that will need to service DHCP requests.

  • Set service_password which is used for communication between services. If unset, a random password is generated during the initial installation and stored on the controller in ~/.config/bifrost/service_password.

The install process, when executed will either download, or build disk images for the deployment of nodes, and be deployed to the nodes.

If you wish to build an image, based upon the settings, you will need to set create_image_via_dib to true.

If you are running the installation behind a proxy, export the environment variables http_proxy, https_proxy and no_proxy so that ansible will use these proxy settings.

TLS support

Bifrost supports TLS for API services with two options:

  • A self-signed certificate can be generated automatically. Set enable_tls=true and generate_tls=true.


    This is equivalent to the --enable-tls flag of bifrost-cli.

  • Certificate paths can be provided via:


    Path to the TLS certificate (must be world-readable).


    Path to the private key (must not be password protected).


    Path to the certificate signing request file.

    Set enable_tls=true and do not set generate_tls to use this option.


If using Keystone, see TLS notes for important notes.


In order to really get started, you must install dependencies.

With the addition of ansible collections, the will install the collections in the default ansible collections_paths (according to your ansible.cfg) or you can specify the location setting ANSIBLE_COLLECTIONS_PATHS:

$ export ANSIBLE_COLLECTIONS_PATHS=/mydir/collections


If you are using a virtual environment ANSIBLE_COLLECTIONS_PATHS is automatically set. After Ansible Collections are installed, a symbolic link to to the installation is created in the bifrost playbook directory.

The script automatically invokes and creates a virtual environment for you:

$ bash ./scripts/
$ source /opt/stack/bifrost/bin/activate
$ cd playbooks

Once the dependencies are in-place, you can execute the ansible playbook to perform the actual installation. The playbook will install and configure ironic in a stand-alone fashion.

A few important notes:

  • The OpenStack Identity service (keystone) is NOT installed by default, and ironic’s API is accessible without authentication. It is possible to put basic password authentication on ironic’s API by changing the nginx configuration accordingly.


Bifrost playbooks can leverage and optionally install keystone. See Keystone install details.

  • The OpenStack Networking service (neutron) is NOT installed. Ironic performs static IP injection via config-drive or DHCP reservation.

  • Deployments are performed by the ironic python agent (IPA).

  • dnsmasq is configured statically and responds to all PXE boot requests by chain-loading to iPXE, which then fetches the Ironic Python Agent ramdisk from nginx.

  • By default, installation will build an Ubuntu-based image for deployment to nodes. This image can be easily customized if so desired.

The re-execution of the playbook will cause states to be re-asserted. If not already present, a number of software packages including MySQL will be installed on the host. Python code will be reinstalled regardless if it has changed.

Playbook Execution

Playbook based install provides a greater degree of visibility and control over the process and is suitable for advanced installation scenarios.


First, make sure that the virtual environment is active (the example below assumes that bifrost venv is installed into the default path /opt/stack/bifrost).

$ . /opt/stack/bifrost/bin/activate (bifrost) $

Verify if the ansible-playbook executable points to the one installed in the virtual environment:

(bifrost) $ which ansible-playbook /opt/stack/bifrost/bin/ansible-playbook (bifrost) $

change to the playbooks subdirectory of the cloned bifrost repository:

$ cd playbooks

If you have passwordless sudo enabled, run:

$ ansible-playbook -vvvv -i inventory/target install.yaml

Otherwise, add the -K to the ansible command line, to trigger ansible to prompt for the sudo password:

$ ansible-playbook -K -vvvv -i inventory/target install.yaml

With regard to testing, ironic’s node cleaning capability is enabled by default, but only metadata cleaning is turned on, as it can be an unexpected surprise for a new user that their test node is unusable for however long it takes for the disks to be wiped.

If you wish to enable full cleaning, you can achieve this by passing the option -e cleaning_disk_erase=true to the command line or executing the command below:

$ ansible-playbook -K -vvvv -i inventory/target install.yaml -e cleaning_disk_erase=true

If installing a stable release, you need to set two more parameters, e.g.:

-e git_branch=stable/train -e ipa_upstream_release=stable-train


Note the difference in format: git branch uses slashes, IPA release uses dashes.

After you have performed an installation, you can edit /etc/ironic/ironic.conf to enable or disable cleaning as desired. It is highly encouraged to utilize cleaning in any production environment.

Additional ironic drivers

An additional collection of drivers are maintained outside of the ironic source code repository, as they do not have Continuous Integration (CI) testing.

These drivers and information about them can be found in ironic-staging-drivers docs. If you would like to install the ironic staging drivers, simply pass -e staging_drivers_include=true when executing the install playbook:

$ ansible-playbook -K -vvvv -i inventory/target install.yaml -e staging_drivers_include=true