KVM/QEMU Virtualization Setup on Linux

KVM/QEMU Virtualization Setup on Linux

Sometimes bare metal hits cost or flexibility limits — and then KVM/QEMU comes in. Clients often come with a problem: "We want virtualization, but without vendor lock-in and with full control." We configure KVM/QEMU turnkey, integrating it into your stack: Terraform, Ansible, GitOps, Prometheus. No GUI, only CLI and API — maximum performance and zero licensing costs. Over the past 5 years, we have implemented more than 50 virtualization projects based on KVM — from small labs to production clusters with 200+ virtual machines. Average budget savings compared to commercial hypervisors — about 40% due to the absence of license fees.

Why Choose KVM/QEMU for Production?

Proxmox is convenient, but its updates and licensing (Community vs Enterprise) may not suit strict environments. KVM/QEMU is a Linux kernel module: you get native performance, not an abstraction layer. According to Red Hat documentation, with VirtIO guests run at 95–98% of bare metal. Live migration, CPU balancing via virsh vcpupin, SLO under 99.9% — all available out of the box without additional overlays. For example, in one project we replaced VMware vSphere with KVM — licensing costs dropped 5 times, and performance increased by 12% thanks to direct kernel access.

How to Ensure KVM Fault Tolerance?

For high-availability environments, we use a combination of live migration and shared storage (NFS, Ceph). Locking setup according to the "2-3 hosts + Quorum" scheme via corosync/pacemaker. Average recovery time (RTO) upon node failure — no more than 15 minutes. Backup plan: qcow2 snapshots every 4 hours, full backup once daily to a separate server. For critical VMs, we configure automatic restart on the neighboring host.

What We Do as Part of KVM/QEMU Setup

1. Hardware audit: check support for VT-x/AMD-V, IOMMU, NUMA. Optimize vCPU and memory placement. On a typical server with 2 sockets of 16 cores each and 256 GB RAM, you can host up to 30 medium-load VMs.
2. Installation and configuration: qemu-kvm, libvirt-daemon, bridge-utils, virtinst. Set up default network (NAT or bridge).
3. Network: create bridge interfaces, VLANs (bridge per VLAN), Open vSwitch optionally. Guarantee NetworkPolicy and isolation. Example: for microservice architecture, we set up one bridge per project.
4. Storage: qcow2 with cache=none,io=native, LVM thin provisioning. Connect NFS/Ceph for live migration.
5. VM templates: cloud-init, images of Ubuntu 22.04 LTS, Debian 12, CentOS Stream 9. Everything via virt-install or Terraform.
6. Monitoring: Prometheus + libvirt_exporter, Grafana dashboards. Metrics: CPU steal, disk latency, memory ballooning.
7. Backups: virt-backup or rsync snapshots. RPO — 1 hour, RTO — 15 minutes.

How to Set Up KVM Monitoring?

Install libvirt_exporter, attach it to Prometheus. Grafana has ready-made dashboards. Key metrics:

• kvm_domain_info — guest states
• kvm_domain_cpu_time_seconds — steal time
• kvm_domain_balloon_current_bytes — memory leaks

Alerting by MTTR and SLO. Runbook: if steal exceeds 10% — check CPU pinning or load on neighboring VMs.

Network Bridge for VMs

Create a Linux bridge like this:

# /etc/netplan/01-bridge.yaml (Ubuntu)
network:
  version: 2
  ethernets:
    enp3s0:
      dhcp4: false
  bridges:
    br0:
      interfaces: [enp3s0]
      addresses: [192.168.1.10/24]
      gateway4: 192.168.1.1
      nameservers:
        addresses: [8.8.8.8]
      parameters:
        stp: false
        forward-delay: 0
netplan apply

For VLAN-tagged traffic, create br-vlan300 over enp3s0.300 — each VLAN gets its own bridge, VMs connect to the appropriate one.

Creating a VM via qemu-system and virt-install

# Create VM disk (qcow2 — thin provisioned, supports snapshots)
qemu-img create -f qcow2 /var/lib/libvirt/images/webserver01.qcow2 40G

# Install VM from ISO
virt-install \
  --name webserver01 \
  --ram 4096 \
  --vcpus 4 \
  --cpu host \
  --disk path=/var/lib/libvirt/images/webserver01.qcow2,format=qcow2,bus=virtio \
  --network bridge=br0,model=virtio \
  --os-variant ubuntu22.04 \
  --cdrom /tmp/ubuntu-22.04-server.iso \
  --graphics none \
  --console pty,target_type=serial \
  --extra-args 'console=ttyS0,115200n8'

Key point: --cpu host passes real processor flags to the VM — maximum performance. If VMs need to migrate between hosts with different CPUs, use --cpu qemu64 for compatibility.

bus=virtio and model=virtio — paravirtualized drivers for disk and network. Performance is significantly higher than emulated IDE/e1000 (3–5 times). Guest OS must have VirtIO drivers (in Linux they are built into the kernel, Windows requires separate installation).

Snapshots and Cloning

# Create a snapshot before update
virsh snapshot-create-as webserver01 \
    --name "before-update" \
    --description "Before kernel update" \
    --atomic

# List snapshots
virsh snapshot-list webserver01

# Revert to snapshot
virsh snapshot-revert webserver01 before-update

# Clone VM
virt-clone --original webserver01 \
    --name webserver02 \
    --file /var/lib/libvirt/images/webserver02.qcow2

qcow2 snapshots store only delta from the base image. With a deep snapshot chain, disk performance degrades — merge old snapshots via qemu-img commit. Manage snapshots via libvirt.

Performance and Tuning

For I/O-intensive VMs, enable cache=none,io=native on disks — disables double caching between guest and host. For network loads — multiqueue on virtio-net: number of queues = number of VM vCPUs.

CPU pinning for latency-sensitive workloads:

# Bind VM vCPUs to specific physical cores virsh vcpupin webserver01 0 2 virsh vcpupin webserver01 1 3 

KVM vs Alternatives

What's Included in Our KVM/QEMU Setup Service

• Infrastructure audit: analyze current loads, select CPU/RAM/disks.
• Installation and configuration: KVM, libvirt, bridge, storage, cloud-init.
• Create VM templates: Ubuntu, Debian, CentOS, Windows Server (with VirtIO drivers).
• Integrate monitoring: Prometheus + Grafana dashboards, SLO alerts.
• Documentation: network diagram, recovery runbook, backup plan.
• Team training: 2–3 sessions on managing virsh, virt-manager (optional).

We've been working with KVM for over 5 years, have delivered projects for 50+ companies — from startups to enterprise.

Setup timeline: from 3 to 5 business days — turnkey, with SLA 99.9% guarantee.

Order an audit of your infrastructure — we'll assess the project and propose the optimal solution. Contact us for a consultation.