Overcloud

Note

This documentation is intended as a walk through of the configuration required for a minimal all-in-one overcloud host. If you are looking for an all-in-one environment for test or development, see Automated Setup.

Preparation

Use the bootstrap user described in prerequisites to access the machine.

As described in the overview, we will use a bridge (br0) and a dummy interface (dummy0) for control plane networking. Use the following commands to create them and assign the bridge a static IP address of 192.168.33.3:

sudo ip l add br0 type bridge
sudo ip l set br0 up
sudo ip a add 192.168.33.3/24 dev br0
sudo ip l add dummy0 type dummy
sudo ip l set dummy0 up
sudo ip l set dummy0 master br0

This configuration is not persistent, and must be recreated if the VM is rebooted.

Installation

Follow the instructions in Installation to set up an Ansible control host environment. Typically this would be on a separate machine, but here we are keeping things as simple as possible.

Configuration

Clone the kayobe-config git repository, using the correct branch for the release you are deploying. In this example we will use the stable/yoga branch.

cd <base path>/src
git clone https://opendev.org/openstack/kayobe-config.git -b stable/yoga
cd kayobe-config

This repository is bare, and needs to be populated. The repository includes an example inventory, which should be removed:

git rm etc/kayobe/inventory/hosts.example

Create an Ansible inventory file and add the machine to it. In this example our machine is called controller0. Since this is an all-in-one environment, we add the controller to the compute group, however normally dedicated compute nodes would be used.

etc/kayobe/inventory/hosts
# This host acts as the configuration management Ansible control host. This must be
# localhost.
localhost ansible_connection=local

[controllers]
controller0

[compute:children]
controllers

The inventory directory also contains group variables for network interface configuration. In this example we will assume that the machine has a single network interface called dummy0. We will create a bridge called br0 and plug dummy0 into it. Replace the network interface configuration for the controllers group with the following:

etc/kayobe/inventory/group_vars/controllers/network-interfaces
# Controller interface on all-in-one network.
aio_interface: br0

# Interface dummy0 is plugged into the all-in-one network bridge.
aio_bridge_ports:
  - dummy0

In this scenario a single network called aio is used. We must therefore set the name of the default controller networks to aio:

etc/kayobe/networks.yml
---
# Kayobe network configuration.

###############################################################################
# Network role to network mappings.

# Map all networks to the all-in-one network.

# Name of the network used for admin access to the overcloud
#admin_oc_net_name:
admin_oc_net_name: aio

# Name of the network used by the seed to manage the bare metal overcloud
# hosts via their out-of-band management controllers.
#oob_oc_net_name:

# Name of the network used by the seed to provision the bare metal overcloud
# hosts.
#provision_oc_net_name:

# Name of the network used by the overcloud hosts to manage the bare metal
# compute hosts via their out-of-band management controllers.
#oob_wl_net_name:

# Name of the network used by the overcloud hosts to provision the bare metal
# workload hosts.
#provision_wl_net_name:

# Name of the network used to expose the internal OpenStack API endpoints.
#internal_net_name:
internal_net_name: aio

# List of names of networks used to provide external network access via
# Neutron.
# Deprecated name: external_net_name
# If external_net_name is defined, external_net_names will default to a list
# containing one item, external_net_name.
#external_net_names:
external_net_names:
  - aio

# Name of the network used to expose the public OpenStack API endpoints.
#public_net_name:
public_net_name: aio

# Name of the network used by Neutron to carry tenant overlay network traffic.
#tunnel_net_name:
tunnel_net_name: aio

# Name of the network used to carry storage data traffic.
#storage_net_name:
storage_net_name: aio

# Name of the network used to carry storage management traffic.
#storage_mgmt_net_name:
storage_mgmt_net_name: aio

# Name of the network used to carry swift storage data traffic.
#swift_storage_net_name:

# Name of the network used to carry swift storage replication traffic.
#swift_storage_replication_net_name:

# Name of the network used to perform hardware introspection on the bare metal
# workload hosts.
#inspection_net_name:

# Name of the network used to perform cleaning on the bare metal workload
# hosts
#cleaning_net_name:

###############################################################################
# Network definitions.

<omitted for clarity>

Next the aio network must be defined. This is done using the various attributes described in Network Configuration. These values should be adjusted to match the environment. The aio_vip_address variable should be a free IP address in the same subnet for the virtual IP address of the OpenStack API.

etc/kayobe/networks.yml
<omitted for clarity>

###############################################################################
# Network definitions.

# All-in-one network.
aio_cidr: 192.168.33.0/24
aio_vip_address: 192.168.33.2

###############################################################################
# Network virtual patch link configuration.

<omitted for clarity>

Kayobe will automatically allocate IP addresses. In this case however, we want to ensure that the host uses the same IP address it has currently, to avoid loss of connectivity. We can do this by populating the network allocation file. Use the correct hostname and IP address for your environment.

etc/kayobe/network-allocation.yml
---
aio_ips:
  controller0: 192.168.33.3

The default OS distribution in Kayobe is CentOS. If using an Ubuntu host, set the os_distribution variable in etc/kayobe/globals.yml to ubuntu or rocky if using Rocky Linux..

etc/kayobe/globals.yml
os_distribution: "ubuntu"

Kayobe uses a bootstrap user to create a stack user account. By default, this user is centos on CentOS, rocky on Rocky and ubuntu on Ubuntu, in line with the default user in the official cloud images. If you are using a different bootstrap user, set the controller_bootstrap_user variable in etc/kayobe/controllers.yml. For example, to set it to cloud-user (as seen in MAAS):

etc/kayobe/controllers.yml
controller_bootstrap_user: "cloud-user"

By default, on systems with SELinux enabled, Kayobe will disable SELinux and reboot the system to apply the change. In a test or development environment this can be a bit disruptive, particularly when using ephemeral network configuration. To avoid rebooting the system after disabling SELinux, set disable_selinux_do_reboot to false in etc/kayobe/globals.yml.

etc/kayobe/globals.yml
disable_selinux_do_reboot: false

In a development environment, we may wish to tune some Kolla Ansible variables. Using QEMU as the virtualisation type will be necessary if KVM is not available. Reducing the number of OpenStack service workers helps to avoid using too much memory.

etc/kayobe/kolla/globals.yml
# Most development environments will use nested virtualisation, and we can't
# guarantee that nested KVM support is available. Use QEMU as a lowest common
# denominator.
nova_compute_virt_type: qemu

# Reduce the control plane's memory footprint by limiting the number of worker
# processes to one per-service.
openstack_service_workers: "1"

We can see the changes that have been made to the configuration.

cd <base path>/src/kayobe-config
git status

On branch master
Your branch is up to date with 'origin/master'.

Changes to be committed:
  (use "git restore --staged <file>..." to unstage)
    deleted:    etc/kayobe/inventory/hosts.example

Changes not staged for commit:
  (use "git add <file>..." to update what will be committed)
  (use "git restore <file>..." to discard changes in working directory)
    modified:   etc/kayobe/globals.yml
    modified:   etc/kayobe/inventory/group_vars/controllers/network-interfaces
    modified:   etc/kayobe/kolla/globals.yml
    modified:   etc/kayobe/networks.yml

Untracked files:
  (use "git add <file>..." to include in what will be committed)
    etc/kayobe/inventory/hosts
    etc/kayobe/network-allocation.yml

The git diff command is also helpful. Once all configuration changes have been made, they should be committed to the kayobe-config git repository.

cd <base path>/src/kayobe-config
git add etc/kayobe/inventory/hosts etc/kayobe/network-allocation.yml
git add --update
git commit -m "All in one scenario config"

In a real environment these changes would be pushed to a central repository.

Deployment

We are now ready to perform a deployment.

Activate the Kayobe virtual environment:

cd <base path>/venvs/kayobe
source bin/activate

Activate the Kayobe configuration environment:

cd <base path>/src/kayobe-config
source kayobe-env

Bootstrap the control host:

kayobe control host bootstrap

Configure the overcloud host:

kayobe overcloud host configure

After this command has run, some files in the kayobe-config repository will have changed. Kayobe performs static allocation of IP addresses, and tracks them in etc/kayobe/network-allocation.yml. Normally there may be changes to this file, but in this case we manually added the IP address of controller0 earlier. Kayobe uses tools provided by Kolla Ansible to generate passwords, and stores them in etc/kayobe/kolla/passwords.yml. It is important to track changes to this file.

cd <base path>/src/kayobe-config
git add etc/kayobe/kolla/passwords.yml
git commit -m "Add autogenerated passwords for Kolla Ansible"

Pull overcloud container images:

kayobe overcloud container image pull

Deploy overcloud services:

kayobe overcloud service deploy

Testing

The init-runonce script provided by Kolla Ansible (not for production) can be used to setup some resources for testing. This includes:

  • some flavors

  • a cirros image

  • an external network

  • a tenant network and router

  • security group rules for ICMP, SSH, and TCP ports 8000 and 8080

  • an SSH key

  • increased quotas

For the external network, use the same subnet as before, with an allocation pool range containing free IP addresses:

pip install python-openstackclient
export EXT_NET_CIDR=192.168.33.0/24
export EXT_NET_GATEWAY=192.168.33.3
export EXT_NET_RANGE="start=192.168.33.4,end=192.168.33.254"
source "${KOLLA_CONFIG_PATH:-/etc/kolla}/admin-openrc.sh"
${KOLLA_SOURCE_PATH}/tools/init-runonce

Create a server instance, assign a floating IP address, and check that it is accessible.

openstack server create --image cirros --flavor m1.tiny --key-name mykey --network demo-net demo1
openstack floating ip create public1

The floating IP address is displayed after it is created, in this example it is 192.168.33.4:

openstack server add floating ip demo1 192.168.33.4
ssh cirros@192.168.33.4