Advanced features


Advanced usage sometimes requires partial execution of the ospd playbook. This can be achieved with Ansible tags

List the available tags of the ospd playbooks:

ir-installer [...] ospd [...] --ansible-args list-tags

Execute only the desired tags. For example, this will only install the UnderCloud and download OverCloud images:

ir-installer [...] ospd [...] --ansible-args "tags=undercloud,images"


Commonly used tags:

Install the UnderCloud.
Download OverCloud images and upload them to UnderCloud’s Glance service.
Create instackenv.json file and perform introspection on OverCloud nodes with Ironic.
Tag Ironic nodes with OverCloud properties.
Generate heat-templates from user provided deployement-files and from input data. Create the accordingly.
Execute script
Do overcloud_init and overcloud_deploy.
Update Ansible inventory and SSH tunneling with new OverCloud nodes details (user, password, keys, etc...)

Common use case of tags is to stop after a certain stage is completed. To do this, Ansible requires a list of all the tags up to, and including the last desired stage. Therefore, in order to stop after UnderCloud is ready:

ir-installer [...] ospd [...] --ansible-args --ansible-args tags="init,dump_facts,undercloud"

Or, in as a bash script, to stop after $BREAKPOINT:

LEADING=`echo $FULL_TAG_LIST | awk -F$BREAKPOINT '{print $1}'`
ir-installer [...] ospd [...] --ansible-args --ansible-args tags=${LEADING}${BREAKPOINT}

OverCloud Image Update

OSPD creates the OverCloud nodes from images. These images should be recreated on any new core build. However, this is not always the case. To updates that image’s packages (after download and before deploying the Overcloud), to match RH-OSP core bits build, set --images-update to yes:

ir-installer [...] ospd [...] --images-update=yes


This might take a while and sometimes hangs. Probably due to old libguestfs packages in RHEL 7.2. For a more detailed console output of that task, set --images-update to verbose.

Custom repositories

Infrared allows to add custom repositories to the UnderCloud when you’re running OSPD, after installing the default repositories of the OSPD release. This can be done passing through --extra-vars with the following key:

  • ospd.extra_repos.from_url which will download a repo file to /etc/yum.repos.d
  1. Using ospd.extra_repos.from_url:

    Create a yaml file:


    Run ir-installer:

    ir-installer --extra-vars=@repos.yml ospd
  1. Using ospd.extra_repos.from_config

    Using this option enables you to set specific options for each repository:

                - { name: my_repo1, file: my_repo1.file, description: my repo1, base_url:, enabled: 0, gpg_check: 0 }
                - { name: my_repo2, file: my_repo2.file, description: my repo2, base_url:, enabled: 0, gpg_check: 0 }


    As you can see, ospd.extra_repos.explicity support some of the options found in yum_repository module (name, file, description, base_url, enabled and gpg_check). For more information about this module, visit Ansible yum_repository documentation.

    Run ir-installer:

    ir-installer -e @repos.yml ospd

Custom/local tempest tester

You might have a specific version of tempest to test locally in a particular directory, and you want to use it. Infrared allows you to use this instead of the default git repository. To do so, all you need to do is pass the key tester.local_dir as extra-vars to ir-tester:

Run ir-tester:

ir-tester tempest --extra-vars tester.local_dir-/patch/for/your/tempest


Infrared allows to perform scale tests on different services.

Currently supported services for tests:
  • compute
  • ceph-storage
  • swift-storage
  1. To scale compute service:

    Deployment should have at least 3 compute nodes.

    Run ansible playbook:

    ansible-playbook -vvvv -i hosts -e @install.yml playbooks/installer/ospd/post_install/scale_compute.yml

    It will scale compute nodes down to 1 and after that scale compute node back to 3.

  2. To scale ceph-storage service:

    Deployment should have at least 3 ceph-storage nodes.

    Run ansible playbook:

    ansible-playbook -vvvv -i hosts -e @install.yml playbooks/installer/ospd/post_install/ceph_compute.yml

    It will scale compute nodes down to 1 and after that scale compute node back to 3.

  3. To scale swift-storage service:

    Deployment should have at least 3 swift-storage nodes.

    Run ansible playbook:

    ansible-playbook -vvvv -i hosts -e @install.yml playbooks/installer/ospd/post_install/swift_compute.yml


    Swift has a parameter called min_part_hours which configures amount of time that should be passed between two rebalance processes. In real production environment this parameter is used to reduce network load. During the deployment of swift cluster for further scale testing we set it to 0 to decrease amount of time for scale.

UnderCloud testing

Usually, all tempest tests are run from the UnderCloud, against OverCloud, while you might want test UnderCloud services (e.g. ironic). The following cookbook uses InfraRed to run Tempest tests against the UnderCloud.

  1. We want an explicit “tester” node to avoid running tests on the same node as the UnderCloud.

    Use “ironic” node instead of “undercloud”. It’s the same but doesn’t have the role of “tester”. Rename “controller” node into “test-vm” to avoid misunderstanding and update it’s parameters to match with “baremetal” flavor.:

    ir-provisioner -d virsh -v -o provision.yml \
        --topology-nodes=ironic:1,controller:1,tester:1 \
        --host-address=$HOST \
        --host-key=$HOME/.ssh/rhos-jenkins/id_rsa \
        --image=$IMAGE \
        -e @private.yml \
        -e provisioner.topology.nodes.controller.cpu=1 \
        -e provisioner.topology.nodes.controller.disks.disk1.size=41G \
        -e provisioner.topology.nodes.controller.memory=4096 \
  2. As we don’t want the full OSPD installation, we will use explicit Tags to do only certain parts:
    • Undercloud - will install UnderCloud

    • Images - installs or builds OverCloud images

    • Ironic - performs all required actions before introspection (including assignment of the kernel and ramdisk)

    • Virthost - enables “virthost” specific tasks in case of “virsh” provisioning:

      ir-installer --debug ospd -v --inventory hosts \
          -e @provision.yml \
          -e @private.yml \
          -o install.yml \
          --deployment-files=$PWD/settings/installer/ospd/deployment/virt \
          --product-version=10 \
          --product-core-version=10 \
  3. We want prepare environment for ironic tests:
    • update baremetal flavor with cpu_arch

    • create initial tempest.conf file using predefined template

    • enable ironic inspector

    • enable fake and pxe_ssh drivers in ironic

    • make desired neutron network shared

    • install rhos-release repos into “tester” node

    • configure data network on “tester” node:

      ansible-playbook -vvvv -i hosts -e @install.yml \
          playbooks/installer/ospd/post_install/add_nodes_to_ironic_list.yml \
          -e net_name=ctlplane \
          -e driver_type=pxe_ssh \
          -e rc_file_name=stackrc
  4. Finally run the Ironic tempest plugin tests::

    Run ir-tester:

    ir-tester --debug tempest -v \
        -e @install.yml \
        --tests=ironic_inspector \
        -o test.yml

Virthost packages/repo requirements