Undercloud Installation

This section contains instructions on how to install the undercloud. For update or upgrade to a deployed undercloud see undercloud_upgrade.

Installing the Undercloud


Instack-undercloud was deprecated in Rocky cycle. Containerized undercloud should be installed instead. See Containers based Undercloud Deployment for backward compatibility related information.


Please ensure all your nodes (undercloud, compute, controllers, etc) have their internal clock set to UTC in order to prevent any issue with possible file future-dated timestamp if hwclock is synced before any timezone offset is applied.

  1. Log in to your machine (baremetal or VM) where you want to install the undercloud as a non-root user (such as the stack user):

    ssh <non-root-user>@<undercloud-machine>


    If you don’t have a non-root user created yet, log in as root and create one with following commands:

    sudo useradd stack
    sudo passwd stack  # specify a password
    echo "stack ALL=(root) NOPASSWD:ALL" | sudo tee -a /etc/sudoers.d/stack
    sudo chmod 0440 /etc/sudoers.d/stack
    su - stack


    The undercloud is intended to work correctly with SELinux enforcing. Installations with the permissive/disabled SELinux are not recommended. The undercloud_enable_selinux config option controls that setting.


    vlan tagged interfaces must follow the if_name.vlan_id convention, like for example: eth0.vlan100 or bond0.vlan120.


    Ensure that there is a FQDN hostname set and that the $HOSTNAME environment variable matches that value. The easiest way to do this is to set the undercloud_hostname option in undercloud.conf before running the install. This will allow the installer to configure all of the hostname- related settings appropriately.

    Alternatively the hostname settings can be configured manually, but this is strongly discouraged. The manual steps are as follows:

    sudo hostnamectl set-hostname myhost.mydomain
    sudo hostnamectl set-hostname --transient myhost.mydomain

    An entry for the system’s FQDN hostname is also needed in /etc/hosts. For example, if the system is named myhost.mydomain, /etc/hosts should have an entry like:   myhost.mydomain myhost
  2. Enable needed repositories:


    Enable optional repo for RHEL7:

    sudo yum install -y yum-utils
    sudo yum-config-manager --enable rhelosp-rhel-7-server-opt


    Python3 is required for current releases of OpenStack which is supported on CentOS Stream 9.

    1. Download and install the python-tripleo-repos RPM from the appropriate RDO repository

      CentOS Stream 9

      Current Centos 9 RDO repository.

      sudo dnf install -y https://trunk.rdoproject.org/centos9/component/tripleo/current/python3-tripleo-repos-<version>.el9.noarch.rpm


      tripleo-repos removes any repositories that it manages before each run. This means all repositories must be specified in a single tripleo-repos call. As an example, the correct way to install the current and ceph repos is to run tripleo-repos current ceph, not two separate calls.

    1. Run tripleo-repos to install the appropriate repositories. The option below will enable the latest master TripleO packages, the latest promoted packages for all other OpenStack services and dependencies and the latest stable Ceph packages. There are other repository configurations available in tripleo-repos, see its --help output for details.

      sudo -E tripleo-repos current-tripleo-dev ceph
  3. Install the TripleO CLI, which will pull in all other necessary packages as dependencies:

    sudo dnf install -y python*-tripleoclient

    RHEL7 / CentOS

    For RHEL or CentOS 7 the command would be:

    sudo yum install -y python-tripleoclient


    If you intend to deploy Ceph in the overcloud, or configure the overcloud to use an external Ceph cluster, and are running Pike or newer, then install ceph-ansible on the undercloud:

    sudo dnf install -y ceph-ansible


    If you intend to deploy TLS-everywhere in the overcloud and are deploying Train with python3 or Ussuri+, install the following packages:

    sudo yum install -y python3-ipalib python3-ipaclient krb5-devel

    If you’re deploying Train with python2, install the corresponding python2 version of the above packages:

    sudo yum install -y python-ipalib python-ipaclient krb5-devel

    if you intend to use Novajoin to implement TLS-everywhere install the following package:

    sudo yum install -y python-novajoin

    You can find more information about deploying with TLS in the TLS Introduction documentation.

  4. Prepare the configuration file:

    cp /usr/share/python-tripleoclient/undercloud.conf.sample ~/undercloud.conf

    It is backwards compatible with non-containerized instack underclouds.

    Stable Branch

    For a non-containerized undercloud, copy in the sample configuration file and edit it to reflect your environment:

    cp /usr/share/instack-undercloud/undercloud.conf.sample ~/undercloud.conf


    There is a tool available that can help with writing a basic undercloud.conf: Undercloud Configuration Wizard It takes some basic information about the intended overcloud environment and generates sane values for a number of the important options.

  5. (OPTIONAL) Generate configuration for preparing container images

    As part of the undercloud install, an image registry is configured on port 8787. This is used to increase reliability of overcloud image pulls, and minimise overall network transfers. The undercloud registry will be populated with images required by the undercloud by generating the following containers-prepare-parameter.yaml file and including it in undercloud.conf: container_images_file=$HOME/containers-prepare-parameter.yaml:

    openstack tripleo container image prepare default \
      --local-push-destination \
      --output-env-file ~/containers-prepare-parameter.yaml


    This command is available since Rocky.

    See Container Image Preparation for details on using containers-prepare-parameter.yaml to control what can be done during the container images prepare phase of an undercloud install.

    Additionally, docker_insecure_registries and docker_registry_mirror parameters allow to customize container registries via the undercloud.conf file.

  6. (OPTIONAL) Override heat parameters and environment files used for undercloud deployment.

    Similarly to overcloud deployments, see Overriding specific templates with local versions and Using a custom location for all templates, the undercloud.conf: custom_env_files and undercloud.conf: templates configuration parameters allow to use a custom heat templates location and override or specify additional information for Heat resources used for undercloud deployment.

    Additionally, the undercloud.conf: roles_file parameter brings in the ultimate flexibility of Deploying with Custom Roles and Deploying with Composable Services. This allows you to deploy an undercloud composed of highly customized containerized services, with the same workflow that TripleO uses for overcloud deployments.


    The CLI and configuration interface used to deploy a containerized undercloud is the same as that used by ‘legacy’ non-containerized underclouds. As noted above however mechanism by which the undercloud is actually deployed is completely changed and what is more, for the first time aligns with the overcloud deployment. See the command openstack tripleo deploy --standalone help for details. It normally should not be used directly for undercloud installations.

  7. Run the command to install the undercloud:


    To deploy an undercloud with SSL, see Deploying with SSL.


    Validations guide will be installed and configured during undercloud installation. You can set enable_validations = false in undercloud.conf to prevent that.

    To deploy an undercloud:

    openstack undercloud install


The undercloud is containerized by default as of Rocky.


It’s possible to enable verbose logging with --verbose option.


To install a deprecated instack undercloud, you’ll need to deploy with --use-heat=False option.

Since Rocky, we will run all the OpenStack services in a moby container runtime unless the default settings are overwritten. This command requires 2 services to be running at all times. The first one is a basic keystone service, which is currently executed by tripleoclient itself, the second one is heat-all which executes the templates and installs the services. The latter can be run on baremetal or in a container (tripleoclient will run it in a container by default).

Once the install has completed, you should take note of the files stackrc and undercloud-passwords.conf. You can source stackrc to interact with the undercloud via the OpenStack command-line client. The undercloud-passwords.conf file contains the passwords used for each service in the undercloud. These passwords will be automatically reused if the undercloud is reinstalled on the same system, so it is not necessary to copy them to undercloud.conf.


Heat installer configuration, logs and state is ephemeral for undercloud deployments. Generated artifacts for consequent deployments get overwritten or removed (when undercloud.conf: cleanup = true). Although, you can still find them stored in compressed files.

Miscellaneous undercloud deployment artifacts, like processed heat templates and compressed files, can be found in undercloud.conf: output_dir locations like ~/tripleo-heat-installer-templates.

There is also a compressed file created and placed into the output dir, named as undercloud-install-<TS>.tar.bzip2, where TS represents a timestamp.

Downloaded ansible playbooks and inventory files (see TripleO config-download User’s Guide: Deploying with Ansible) used for undercloud deployment are stored in the tempdir ~/undercloud-ansible-<XXXX> by default.


In order to obtain the ansible command used for the installation of the Undercloud in the artifacts directory, it is necessary to pass the option --reproduce-command in the Undercloud deployment command.


Any passwords set in undercloud.conf will take precedence over the ones in undercloud-passwords.conf.


The undercloud installation command can be rerun to reapply changes from undercloud.conf to the undercloud. Note that this should be done with caution if an overcloud has already been deployed or is in progress as some configuration changes could affect the overcloud. These changes include but are not limited to:

  1. Package repository changes on the undercloud, followed by running the installation command could update the undercloud such that further management operations are not possible on the overcloud until the overcloud update or upgrade procedure is followed.

  2. Reconfiguration of the undercloud container registry if the overcloud is using the undercloud as the source for container images.

  3. Networking configuration changes on the undercloud which may affect the overcloud’s ability to connect to the undercloud for instance metadata services.


If running docker commands as a stack user after an undercloud install fail with a permission error, log out and log in again. The stack user does get added to the docker group during install, but that change gets reflected only after a new login.

Cleaning the Undercloud

This procedure isn’t cleaning everything that TripleO generates, but enough so an Undercloud could be re-deployed.


This procedure has been tested on Train and onward. There is no guarantee that it works before this version, due to container commands and new directories.

  1. Log in to your machine (baremetal or VM) where you want to cleanup the undercloud as a non-root user (such as the stack user):

    ssh <non-root-user>@<undercloud-machine>
  2. Cleanup the containers and their images:

    sudo podman rm -af
    sudo podman rmi -af
  3. Remove directories generated by TripleO:

    sudo rm -rf \
        /var/lib/tripleo-config \
        /var/lib/config-data \
        /var/lib/container-config-scripts \
        /var/lib/container-puppet \
        /var/lib/heat-config \
        /var/lib/image-service \
  4. Cleanup systemd:

    sudo rm -rf /etc/systemd/system/tripleo*
    sudo systemctl daemon-reload