OpenStack Deployment With Kolla Ansible | High Availability

This project documents the complete process of deploying a highly available (HA) OpenStack cloud environment using Kolla-Ansible. The deployment includes:

• Multi-node setup (controller, compute, and storage nodes)
• Full containerization using Docker
• Galera cluster, HAProxy, and Keepalived for HA
• Centralized configuration using Ansible
• Virtual machines provisioned with KVM/QEMU

The guide is designed to help DevOps engineers, cloud architects, and advanced system administrators replicate the deployment in a lab or production-ready setup.

Tested Environment:

🖥️ OS: Rocky Linux 9
☁️ Cloud Stack: OpenStack 2024.2
⚙️ Tools: Kolla-Ansible, Docker, Ansible

📘 Note: This documentation assumes a basic understanding of Linux, networking, and Ansible.

---

🧱 Environment Overview

• The deployment was performed in a virtualized lab using VMWare Workstation.
• One base virtual machine (VM) was created and then cloned to spawn additional nodes with the appropriate roles assigned.
• All VMs run Rocky Linux 9.4 and share the same hardware specs initially, with modifications based on the role.

📥 Download Rocky Linux 9.4 ISO:
https://rockylinux.org/download

---

🧩 Roles of Each Node in the OpenStack Deployment

This section describes the specific responsibilities and key services hosted on each node type within the HA OpenStack cluster.

🧠 Controller Node

Primary Role:
The "brain" of the OpenStack cluster.

Key Services:

• Keystone: Identity management
• Glance: Image storage for VMs
• Nova API: Compute API endpoints
• Horizon: Web dashboard interface
• MariaDB/Galera: Highly available database cluster
• RabbitMQ: Message queuing between OpenStack services

Responsibility:
Handles orchestration, authentication, database operations, and API endpoints for all OpenStack components.

---

🧮 Compute Node

Primary Role:
Runs virtual machines (VMs) and provides compute resources.

Key Services:

• Nova Compute: Manages VM lifecycle (create, delete, migrate, etc.)
• Libvirt/KVM: Hypervisor interface to run VMs

Responsibility:
Provides vCPUs, RAM, and local storage for VMs. Compute nodes scale horizontally — adding more nodes increases cluster capacity.

---

🌐 Network Node

Primary Role:
Manages all networking components of the OpenStack environment.

Key Services:

• Neutron: Networking service (routers, subnets, floating IPs)
• L3 Agent: Routing and NAT for VM networks
• DHCP Agent: Dynamic IP assignment to VMs
• OVS / LinuxBridge: Virtual switch to connect instances

Responsibility:
Ensures full VM connectivity, external access, router/NAT functionality, and applies security group rules.

---

💾 Storage Node

Primary Role:
Provides persistent storage to VMs (and optionally object storage).

Key Services:

• Cinder Volume: Block storage for VMs using LVM as the backend
• (Optional) Swift: Object storage (not enabled in this deployment)

Responsibility:
Manages volumes, snapshots, and backups. In our setup, LVM aggregates additional disks into a single volume group (cinder-volumes) for dynamic block provisioning.

---

🖥️ Deployment Architecture

This section outlines the architecture of the OpenStack HA deployment, including the roles assigned to each node, their hardware specifications, and how they are organized to ensure scalability, high availability, and separation of concerns.

Hostname	Role	IPv4	vCPU	RAM (GB)	Storage (GB)	Notes
controller01	Controller Node	192.168.142.141	2	8	40	Also used to deploy Kolla
controller02	Controller Node	192.168.142.142	2	8	40
controller03	Controller Node	192.168.142.143	2	8	40
compute01	Compute Node	192.168.142.151	2	8	40	Virtualization enabled
compute02	Compute Node	192.168.142.152	2	8	40	Virtualization enabled
network01	Network Node	192.168.142.161	2	8	40
network02	Network Node	192.168.142.162	2	8	40
storage01	Storage (Cinder LVM)	192.168.142.171	2	8	40 (+10 GB)	LVM volume for Cinder
storage02	Storage (Cinder LVM)	192.168.142.172	2	8	40 (+10 GB)	LVM volume for Cinder

📌 Note: Each VM was cloned from controller01 and then customized (hostname, static IP, NICs, etc.).

---

🔧 VM Networking

Each VM includes at least two network interfaces:

• ens160: Internal/OpenStack management network
• ens192: External network for floating IPs and external access

NIC names may vary depending on the hypervisor configuration.

⚠️ Reminder: Always confirm NIC names using ip a after VM creation.

⚙️ VM Creation and OS Configuration

This section covers the step-by-step creation of the base Rocky Linux VM, cloning it to form the OpenStack cluster, OS preparation, static IP setup, and SSH key configuration.

🖥️ Creating and Installing Base Rocky Linux VM

1. Create a virtual machine named controller01 which will serve as the base image.
   It will be cloned later to create other nodes, and each clone will be customized.
2. Create the VM using the previously specified hardware (e.g., 2 vCPUs, 8 GB RAM, 40 GB disk).

3. Add another NIC to the VM to support internal and external network separation.

4. Confirm the final specifications before finishing the setup.

5. Start the VM and begin installing Rocky Linux 9.4.

6. Choose your preferred installation language.

7. Configure installation settings:
   • Partitioning
   • Networking
   • Time zone
   • User creation

8. Create a user named kolla to be used for Kolla Ansible deployment.

9. Network Configuration:

• Keep DHCP enabled on the first NIC (we’ll assign a static IP later).
• Disable IPv4 on the second NIC (e.g., ens192).

• Disable → Enable the NIC to apply settings.

10. Finish configuration and begin the OS installation.

11. After reboot, log in as root and update the system:

sudo dnf update -y

12. Add the kolla user to the wheel group:

usermod -aG wheel kolla
grep wheel /etc/group

13. Edit the sudoers file to allow passwordless sudo for the wheel group:

visudo

• Find:

%wheel  ALL=(ALL)       ALL

• Comment that line and uncomment:

%wheel  ALL=(ALL)       NOPASSWD: ALL

---

📦 Cloning the Base VM

1. Shut down the base VM (controller01) before cloning.
2. Use full independent clone option.

3. Repeat the process to create the full architecture:
   • controller02, controller03, compute01, compute02, network01, network02, storage01, storage02.

---

🧾 Configure Hostname and Static IP for All VMs

Each VM needs a unique hostname and static IP.

1. Set the hostname:

   sudo hostnamectl set-hostname <hostname>

2. Validate:

   hostname

3. Set static IP for ens160:

   nmcli device status

sudo nmcli con mod ens160 ipv4.addresses 192.168.142.141/24
sudo nmcli con mod ens160 ipv4.gateway 192.168.142.2
sudo nmcli con mod ens160 ipv4.dns 192.168.142.2
sudo nmcli con mod ens160 ipv4.method manual

sudo nmcli con down ens160 && sudo nmcli con up ens160
ip a show ens160

Repeat for each VM with appropriate IP and hostname.

---

📁 Configure /etc/hosts on controller01

• This node will act as the Kolla-Ansible deployment host.
• This file ensures all nodes in the cluster can be referenced by hostname during the Kolla Ansible deployment.
• DNS entries in other nodes will be taken care by Ansible.

1. Open the hosts file:

   sudo nano /etc/hosts

2. Add these entries:

   192.168.142.141 controller01
   192.168.142.142 controller02
   192.168.142.143 controller03
   192.168.142.151 compute01
   192.168.142.152 compute02
   192.168.142.161 network01
   192.168.142.162 network02
   192.168.142.171 storage01
   192.168.142.172 storage02
   

3. Test reachability:

   for host in controller01 controller02 controller03 compute01 compute02 network01 network02 storage01 storage02; do
     ping -c 1 $host >/dev/null && echo "$host is reachable" || echo "$host is NOT reachable"
   done

---

🧬 Enable Virtualization on Compute VMs

1. Open VM settings for each compute node.

2. Enable virtualization in the hardware settings.

3. Repeat for compute02.

---

💽 Attach New Disk to Storage VMs (for Cinder LVM)

1. Add a new virtual disk (20GB or more) to each storage node.

2. Verify:

   lsblk

3. Initialize and configure LVM:

   sudo pvcreate /dev/nvme0n2
   sudo vgcreate cinder-volumes /dev/nvme0n2
   sudo vgs

4. Repeat on storage02.

---

🔐 Generate SSH Keys for kolla User

Now, we will generate an SSH key for the kolla user on controller01 and copy the public key to all other nodes under the same user (kolla). This is essential for password less SSH access, which Kolla Ansible uses to operate on remote nodes.

1. Log in as kolla on controller01.

2. Generate key pair:

   ssh-keygen -t rsa -b 4096 -N "" -f ~/.ssh/id_rsa

3. Copy public key to all nodes:

   for host in controller01 controller02 controller03 compute01 compute02 network01 network02 storage01 storage02; do
     ssh-copy-id kolla@$host
   done

📌 You’ll be prompted for the kolla user password on each host.

4. Test:

   ssh kolla@controller03

---

🚀 Installing OpenStack with Kolla-Ansible

All the following steps are executed on the deployment node (controller01), which orchestrates the entire OpenStack environment using Kolla Ansible.

---

📦 Install Dependencies

1. Update system packages:

   sudo dnf update -y

2. Install Python build dependencies:

   sudo dnf install git python3-devel libffi-devel gcc openssl-devel python3-libselinux -y

3. Install Ansible:

   sudo pip3 install ansible-core

⚠️ If you face a PATH warning, update your shell configuration:

nano ~/.bashrc
# Add this line
export PATH="/usr/local/bin:$PATH"
source ~/.bashrc

---

🧰 Install Kolla Ansible

1. Install kolla-ansible via pip:

   sudo pip3 install kolla-ansible

2. Create the configuration directory:

   sudo mkdir -p /etc/kolla
   sudo chown $USER:$USER /etc/kolla

3. Copy example config files:

   cp -r /usr/local/share/kolla-ansible/etc_examples/kolla/* /etc/kolla/

4. Copy the multinode inventory file to the current directory:

   cp /usr/local/share/kolla-ansible/ansible/inventory/multinode .

   ℹ️ We're using multinode to deploy a highly available OpenStack setup across multiple nodes.

5. Install Ansible Galaxy dependencies:

   kolla-ansible install-deps

---

⚙️ Configure Main Files

globals.yml

• Path: /etc/kolla/globals.yml
• Purpose: Main config file for customizing OpenStack deployment.

nano /etc/kolla/globals.yml

Paste the following configuration:

🔽 Click to Expand `globals.yml` Example

##########################
# General Configuration
##########################
config_strategy: "COPY_ALWAYS"
kolla_base_distro: "rocky"
openstack_release: "2024.2"
network_address_family: "ipv4"
kolla_container_engine: docker
docker_configure_for_zun: "yes"
containerd_configure_for_zun: "yes"
docker_apt_package_pin: "5:20.*"

##########################
# Network Configuration
##########################
network_interface: "ens160"
neutron_external_interface: "ens192"
kolla_internal_vip_address: "192.168.142.250"

##########################
# OpenStack Core Services
##########################
enable_openstack_core: "yes"
enable_keystone: "{{ enable_openstack_core | bool }}"
enable_glance: "{{ enable_openstack_core | bool }}"
enable_nova: "{{ enable_openstack_core | bool }}"
enable_neutron: "{{ enable_openstack_core | bool }}"
enable_heat: "{{ enable_openstack_core | bool }}"
enable_horizon: "{{ enable_openstack_core | bool }}"

##########################
# Networking Plugin
##########################
neutron_plugin_agent: "openvswitch"
enable_kuryr: "yes"

##########################
# High Availability
##########################
enable_haproxy: "yes"
enable_keepalived: "{{ enable_haproxy | bool }}"

##########################
# Database & Messaging
##########################
enable_mariadb: "yes"
enable_memcached: "yes"
enable_etcd: "yes"

##########################
# Telemetry & Monitoring
##########################
enable_ceilometer: "yes"
enable_aodh: "yes"
enable_gnocchi: "yes"
enable_gnocchi_statsd: "yes"
enable_prometheus: "yes"
enable_grafana: "yes"

##########################
# Block Storage (Cinder)
##########################
enable_cinder: "yes"
enable_cinder_backend_lvm: "yes"
cinder_volume_group: "cinder-volumes"

##########################
# Containerized Application Services
##########################
enable_zun: "yes"
enable_horizon_zun: "{{ enable_zun | bool }}"

##########################
# Image Service Configuration (Glance)
##########################
glance_backend_file: "yes"
glance_file_datadir_volume: "/mnt/glance"

---

passwords.yml

• Path: /etc/kolla/passwords.yml

• Purpose: Stores auto-generated or custom passwords for all OpenStack services.

• Generate it with:

  kolla-genpwd

---

multinode Inventory File

• Defines node roles and groups for the Ansible deployment.

[control]
controller01
controller02
controller03

[network]
network01
network02

[compute]
compute01
compute02

[monitoring]
controller01

[storage]
storage01
storage02

[all:vars]
ansible_user=kolla
ansible_become=True

[deployment]
localhost ansible_connection=local

---

📤 Deploy OpenStack

1. Bootstrap the servers:

   kolla-ansible bootstrap-servers -i ./multinode

2. Run pre-deployment checks:

   kolla-ansible prechecks -i ./multinode

3. Deploy OpenStack:

   kolla-ansible deploy -i ./multinode

4. Validate service configurations:

   kolla-ansible validate-config -i ./multinode

---

🧪 Using OpenStack After Deployment

1. Post-deployment config (generates clouds.yaml):

   kolla-ansible post-deploy

2. Verify clouds.yaml file:

   ls /etc/kolla/clouds.yaml

3. Install OpenStack CLI:

   pip install python-openstackclient -c https://releases.openstack.org/constraints/upper/master

4. Test the deployment:

• Try to list the services:

openstack service list

• Source OpenStack admin credentials file (admin-openrc.sh) that allows you to interact with the OpenStack CLI

source /etc/kolla/admin-openrc.sh

• Check the status of compute nodes:

openstack compute service list

5. Access Horizon Dashboard: Open your browser and visit: http://192.168.142.250

   Login credentials:

   • Username: admin

   • Password: (find it using)

     grep keystone_admin_password /etc/kolla/passwords.yml

---

🧪 (Optional) Setup Demo Environment

/usr/local/share/kolla-ansible/init-runonce

---

🖧 Explore Running Services on Nodes

Example Services by Role:

Storage Node (storage02)

• cinder_backup, cinder_volume, iscsid
• prometheus_cadvisor, fluentd

Network Node (network01)

• neutron_l3_agent, neutron_dhcp_agent, neutron_openvswitch_agent
• keepalived, haproxy

✅ Conclusion

This project demonstrates a full deployment of a highly available OpenStack cloud using Kolla-Ansible and Docker containers across multiple virtualized nodes. The setup includes key OpenStack services distributed among controller, compute, network, and storage nodes, all integrated with high availability components like Galera, HAProxy, and Keepalived.

By following this guide, system administrators and DevOps engineers can:

• Understand the architecture of a production-grade OpenStack environment
• Reproduce a working HA deployment for testing, learning, or production use
• Customize the setup further to include services like Swift, Ceilometer, or Barbican

🚀 Next Steps:
Consider integrating monitoring tools (Prometheus/Grafana), backups, and automation scripts to manage and scale the cloud environment more efficiently.