Migrating from VMware to oVirt: Why Now?
Since Broadcom acquired VMware and tightened licensing policies, sysadmins have been looking for alternatives. Many choose Proxmox VE — I even run a homelab cluster with 12 VMs for testing. However, for large-scale production environments with hundreds of nodes, strict high availability (HA) requirements, and centralized storage management, oVirt is truly the solution capable of replacing vSphere.
In essence, oVirt is the upstream project for Red Hat Virtualization (RHV). If you are familiar with the RHEL, CentOS, or AlmaLinux ecosystems, you will find oVirt very intuitive. After more than 6 months of running oVirt to replace old VMware servers, I’ve found that it handles centralized KVM management smoothly and, most importantly, saves thousands of dollars in licensing fees annually.
Infrastructure Preparation: Don’t Skimp on Hardware
To run oVirt stably, you need to invest in decent resources. Unlike standalone KVM, the oVirt Engine management “brain” is quite resource-heavy. Here is the actual configuration I recommend for a node running both the Engine and VMs (Hosted Engine):
- CPU: Minimum 8 cores (Must support hardware virtualization VT-x or AMD-V).
- RAM: 16GB or more. In practice, the Engine alone takes 4GB to ensure the dashboard doesn’t lag.
- Disk: 100GB for OS; SSD or NVMe is mandatory to avoid I/O bottlenecks when starting many VMs simultaneously.
- Network: At least two 1Gbps ports. If using network storage (NFS/iSCSI), prioritize 10Gbps cards.
- OS: AlmaLinux 8 or CentOS Stream 8 are currently the most stable platforms for oVirt 4.5.
A critical note: You must configure the FQDN (Fully Qualified Domain Name) correctly. oVirt is extremely sensitive to DNS resolution. Even a single typo in the /etc/hosts file can cause the installation script to hang at 80%.
Deploying oVirt Engine as a “Hosted Engine”
Instead of installing the Engine on a separate physical machine, I recommend using the “Hosted Engine” approach. This allows the Engine to run inside a special VM on the virtualization nodes, saving server space and providing self-healing capabilities (Self-Hosted HA) if the primary node fails.
Start by adding the repository and installing the setup package:
# Add oVirt 4.5 repository
sudo dnf install -y https://resources.ovirt.org/pub/yum-repo/ovirt-release45.rpm
# Update the system to the latest version
sudo dnf update -y
# Install the deployment tool
sudo dnf install -y ovirt-hosted-engine-setupNext, run the deploy command. This is when you need to focus carefully to answer the configuration prompts:
sudo hosted-engine --deployThis process will automatically download the image, create the VM, and configure the entire system. You need to prepare a dedicated static IP for the Engine. For production environments, prioritize NFS or iSCSI for Storage to enable Live Migration later.
Storage and Networking Configuration: Key Points to Remember
1. Storage Management
In oVirt, a Storage Domain is where virtual disks and ISO files are stored. A practical tip is to separate storage for Data (VM disks) and ISOs. If using NFS, remember to grant write permissions to the vdsm user (UID 36) on the storage server. If this step is missed, oVirt will report a mount error immediately.
2. Network Management
oVirt uses Linux Bridge to coordinate traffic. You can separate Logical Networks such as: Management, Migration, and VM Network (application traffic). If the system uses VLANs, you simply assign the VLAN ID in the Dashboard and drag-and-drop it onto the physical network card. This intuitive method is much faster than manually typing ip link commands on each machine.
Real-world Evaluation After 6 Months
After six months of running our internal ERP system on oVirt, I’ve identified three major benefits:
- Monitoring capabilities: The built-in Grafana dashboard provides a comprehensive view of Cluster performance. Historical charts make it easy to identify VMs consuming too many resources during peak hours.
- Smooth Live Migration: When Node 1 requires hardware maintenance, I simply switch it to Maintenance mode. oVirt automatically migrates all VMs to Node 2. With a 10Gbps network, latency is almost zero, and users don’t notice any disruption.
- Reliable HA mechanism: Once, a faulty RAM module caused Node 1 to crash unexpectedly. The Engine detected it within seconds and automatically restarted five critical VMs on Node 2. Remember to configure Power Management (Fencing) to prevent “split-brain” scenarios (where two machines fight for data control).
For deeper inspection, you can use the Cockpit tool on port 9090:
# Access the host management dashboard
https://your-host-ip:9090Overall, oVirt is not for those who prefer the simplicity of VirtualBox. It has a steeper learning curve because it is designed for Datacenter scale. But once mastered, you will have a powerful, stable, and completely free virtualization infrastructure capable of handling the heaviest enterprise workloads.
