Deployment

This section describes usage of Kayobe to install an OpenStack cloud onto a set of bare metal servers. We assume access is available to a node which will act as the hypervisor hosting the seed node in a VM. We also assume that this seed hypervisor has access to the bare metal nodes that will form the OpenStack control plane. Finally, we assume that the control plane nodes have access to the bare metal nodes that will form the workload node pool.

Ansible Control Host

Before starting deployment we must bootstrap the Ansible control host. Tasks performed here include:

  • Install Ansible and role dependencies from Ansible Galaxy.
  • Generate an SSH key if necessary and add it to the current user’s authorised keys.

To bootstrap the Ansible control host:

(kayobe) $ kayobe control host bootstrap

Physical Network

The physical network can be managed by Kayobe, which uses Ansible’s network modules. Currently Dell Network OS 6 and Dell Network OS 9 switches are supported but this could easily be extended. To provision the physical network:

(kayobe) $ kayobe physical network configure --group <group> [--enable-discovery]

The --group argument is used to specify an Ansible group containing the switches to be configured.

The --enable-discovery argument enables a one-time configuration of ports attached to baremetal compute nodes to support hardware discovery via ironic inspector.

It is possible to limit the switch interfaces that will be configured, either by interface name or interface description:

(kayobe) $ kayobe physical network configure --group <group> --interface-limit <interface names>
(kayobe) $ kayobe physical network configure --group <group> --interface-description-limit <interface descriptions>

The names or descriptions should be separated by commas. This may be useful when adding compute nodes to an existing deployment, in order to avoid changing the configuration interfaces in use by active nodes.

The --display argument will display the candidate switch configuration, without actually applying it.

Seed Hypervisor

Note

It is not necessary to run the seed services in a VM. To use an existing bare metal host or a VM provisioned outside of Kayobe, this section may be skipped.

Host Configuration

To configure the seed hypervisor’s host OS, and the Libvirt/KVM virtualisation support:

(kayobe) $ kayobe seed hypervisor host configure

Seed

VM Provisioning

Note

It is not necessary to run the seed services in a VM. To use an existing bare metal host or a VM provisioned outside of Kayobe, this step may be skipped. Ensure that the Ansible inventory contains a host for the seed.

The seed hypervisor should have CentOS and libvirt installed. It should have libvirt networks configured for all networks that the seed VM needs access to and a libvirt storage pool available for the seed VM’s volumes. To provision the seed VM:

(kayobe) $ kayobe seed vm provision

When this command has completed the seed VM should be active and accessible via SSH. Kayobe will update the Ansible inventory with the IP address of the VM.

Host Configuration

To configure the seed host OS:

(kayobe) $ kayobe seed host configure

Note

If the seed host uses disks that have been in use in a previous installation, it may be necessary to wipe partition and LVM data from those disks. To wipe all disks that are not mounted during host configuration:

(kayobe) $ kayobe seed host configure --wipe-disks

Building Container Images

Note

It is possible to use prebuilt container images from an image registry such as Dockerhub. In this case, this step can be skipped.

It is possible to use prebuilt container images from an image registry such as Dockerhub. In some cases it may be necessary to build images locally either to apply local image customisation or to use a downstream version of kolla. Images are built by hosts in the container-image-builders group, which by default includes the seed.

To build container images:

(kayobe) $ kayobe seed container image build

It is possible to build a specific set of images by supplying one or more image name regular expressions:

(kayobe) $ kayobe seed container image build bifrost-deploy

In order to push images to a registry after they are built, add the --push argument.

Workaround VLAN cloud-init issue

If you wish to configure the overcloud hosts to use a tagged VLAN for the admin network interface, you must set overcloud_host_image_workaround_cloud_init_enabled to True in ${KAYOBE_CONFIG_PATH}/etc/kayobe/overcloud.yml:

overcloud_host_image_workaround_cloud_init_enabled: True

prior to deploying the containerised services with:

(kayobe) $ kayobe seed service deploy

Kayobe will then patch the overcloud host image to include a more recent version of cloud-init. This is to workaround a bug in the version of cloud-init currently shipped with CentOS 7.5 (0.7.9-24 at the time of writing), which doesn’t set the IP address of VLAN subinterfaces. See: https://bugs.centos.org/view.php?id=14964.

The default repository used to obtain the package is currently hosted on github in the cloud-init-repo repository. You can override this by setting overcloud_host_image_workaround_cloud_init_repo in ${KAYOBE_CONFIG_PATH}/etc/kayobe/overcloud.yml:

overcloud_host_image_workaround_cloud_init_repo: https://stackhpc.github.io/cloud-init-repo/

The source code used to build the updated package can be obtained from the cloud-init-repo-source repository.

As this is not an offical package, there may be latent bugs when using functionality the kayobe developers have not used themselves.

Deploying Containerised Services

At this point the seed services need to be deployed on the seed VM. These services are deployed in the bifrost_deploy container. This command will also build the Operating System image that will be used to deploy the overcloud nodes using Disk Image Builder (DIB).

To deploy the seed services in containers:

(kayobe) $ kayobe seed service deploy

After this command has completed the seed services will be active.

Building Deployment Images

Note

It is possible to use prebuilt deployment images. In this case, this step can be skipped.

It is possible to use prebuilt deployment images from the OpenStack hosted tarballs or another source. In some cases it may be necessary to build images locally either to apply local image customisation or to use a downstream version of Ironic Python Agent (IPA). In order to build IPA images, the ipa_build_images variable should be set to True. To build images locally:

(kayobe) $ kayobe seed deployment image build

If images have been built previously, they will not be rebuilt. To force rebuilding images, use the --force-rebuild argument.

Accessing the Seed via SSH (Optional)

For SSH access to the seed, first determine the seed’s IP address. We can use the kayobe configuration dump command to inspect the seed’s IP address:

(kayobe) $ kayobe configuration dump --host seed --var-name ansible_host

The kayobe_ansible_user variable determines which user account will be used by Kayobe when accessing the machine via SSH. By default this is stack. Use this user to access the seed:

$ ssh <kayobe ansible user>@<seed VM IP>

To see the active Docker containers:

$ docker ps

Leave the seed VM and return to the shell on the Ansible control host:

$ exit

Overcloud

Discovery

Note

If discovery of the overcloud is not possible, a static inventory of servers using the bifrost servers.yml file format may be configured using the kolla_bifrost_servers variable in ${KAYOBE_CONFIG_PATH}/bifrost.yml.

Discovery of the overcloud is supported by the ironic inspector service running in the bifrost_deploy container on the seed. The service is configured to PXE boot unrecognised MAC addresses with an IPA ramdisk for introspection. If an introspected node does not exist in the ironic inventory, ironic inspector will create a new entry for it.

Discovery of the overcloud is triggered by causing the nodes to PXE boot using a NIC attached to the overcloud provisioning network. For many servers this will be the factory default and can be performed by powering them on.

On completion of the discovery process, the overcloud nodes should be registered with the ironic service running in the seed host’s bifrost_deploy container. The node inventory can be viewed by executing the following on the seed:

$ docker exec -it bifrost_deploy bash
(bifrost_deploy) $ source env-vars
(bifrost_deploy) $ ironic node-list

In order to interact with these nodes using Kayobe, run the following command to add them to the Kayobe and Kolla-Ansible inventories:

(kayobe) $ kayobe overcloud inventory discover

Saving Hardware Introspection Data

If ironic inspector is in use on the seed host, introspection data will be stored in the local nginx service. This data may be saved to the control host:

(kayobe) $ kayobe overcloud introspection data save

--output-dir may be used to specify the directory in which introspection data files will be saved. --output-format may be used to set the format of the files.

BIOS and RAID Configuration

Note

BIOS and RAID configuration may require one or more power cycles of the hardware to complete the operation. These will be performed automatically.

Configuration of BIOS settings and RAID volumes is currently performed out of band as a separate task from hardware provisioning. To configure the BIOS and RAID:

(kayobe) $ kayobe overcloud bios raid configure

After configuring the nodes’ RAID volumes it may be necessary to perform hardware inspection of the nodes to reconfigure the ironic nodes’ scheduling properties and root device hints. To perform manual hardware inspection:

(kayobe) $ kayobe overcloud hardware inspect

Provisioning

Provisioning of the overcloud is performed by the ironic service running in the bifrost container on the seed. To provision the overcloud nodes:

(kayobe) $ kayobe overcloud provision

After this command has completed the overcloud nodes should have been provisioned with an OS image. The command will wait for the nodes to become active in ironic and accessible via SSH.

Host Configuration

To configure the overcloud hosts’ OS:

(kayobe) $ kayobe overcloud host configure

Note

If the controller hosts use disks that have been in use in a previous installation, it may be necessary to wipe partition and LVM data from those disks. To wipe all disks that are not mounted during host configuration:

(kayobe) $ kayobe overcloud host configure --wipe-disks

Building Container Images

Note

It is possible to use prebuilt container images from an image registry such as Dockerhub. In this case, this step can be skipped.

In some cases it may be necessary to build images locally either to apply local image customisation or to use a downstream version of kolla. Images are built by hosts in the container-image-builders group, which by default includes the seed. If no seed host is in use, for example in an all-in-one controller development environment, this group may be modified to cause containers to be built on the controllers.

To build container images:

(kayobe) $ kayobe overcloud container image build

It is possible to build a specific set of images by supplying one or more image name regular expressions:

(kayobe) $ kayobe overcloud container image build ironic- nova-api

In order to push images to a registry after they are built, add the --push argument.

Pulling Container Images

Note

It is possible to build container images locally avoiding the need for an image registry such as Dockerhub. In this case, this step can be skipped.

In most cases suitable prebuilt kolla images will be available on Dockerhub. The stackhpc account provides image repositories suitable for use with kayobe and will be used by default. To pull images from the configured image registry:

(kayobe) $ kayobe overcloud container image pull

Building Deployment Images

Note

It is possible to use prebuilt deployment images. In this case, this step can be skipped.

Note

Deployment images are only required for the overcloud when Ironic is in use. Otherwise, this step can be skipped.

It is possible to use prebuilt deployment images from the OpenStack hosted tarballs or another source. In some cases it may be necessary to build images locally either to apply local image customisation or to use a downstream version of Ironic Python Agent (IPA). In order to build IPA images, the ipa_build_images variable should be set to True. To build images locally:

(kayobe) $ kayobe overcloud deployment image build

If images have been built previously, they will not be rebuilt. To force rebuilding images, use the --force-rebuild argument.

Deploying Containerised Services

To deploy the overcloud services in containers:

(kayobe) $ kayobe overcloud service deploy

Once this command has completed the overcloud nodes should have OpenStack services running in Docker containers.

Interacting with the Control Plane

Kolla-ansible writes out an environment file that can be used to access the OpenStack admin endpoints as the admin user:

$ source ${KOLLA_CONFIG_PATH:-/etc/kolla}/admin-openrc.sh

Kayobe also generates an environment file that can be used to access the OpenStack public endpoints as the admin user which may be required if the admin endpoints are not available from the Ansible control host:

$ source ${KOLLA_CONFIG_PATH:-/etc/kolla}/public-openrc.sh

Performing Post-deployment Configuration

To perform post deployment configuration of the overcloud services:

(kayobe) $ source ${KOLLA_CONFIG_PATH:-/etc/kolla}/admin-openrc.sh
(kayobe) $ kayobe overcloud post configure

This will perform the following tasks:

  • Register Ironic Python Agent (IPA) images with glance
  • Register introspection rules with ironic inspector
  • Register a provisioning network and subnet with neutron
  • Configure Grafana organisations, dashboards and datasources