Two-Site High Availability Without the Enterprise Price Tag

A business running a customer-facing web platform needed it to stay online if a server, a building or an internet connection failed, without paying for enterprise load balancers or a cloud bill many times its hosting budget.

The challenge

The two servers sat in different locations, each with its own internet provider and its own public address. Most high-availability designs assume both servers share a network, and these didn’t. If the platform moved to the second site, the internet had to start sending visitors to a completely different address.

What we did

• A two-node cluster: Proxmox at both sites, plus a small witness device so the cluster always knows which side is healthy
• Continuous replication: The platform’s virtual machine copied to the second site every five minutes, and restarted there automatically if the first site fails
• A self-aware server: A failover monitor we wrote inside the virtual machine works out which site it’s running at, switches its network settings to match and updates the public DNS record through Cloudflare
• No false alarms: The monitor tells an internet outage apart from a genuine move, so a flaky connection doesn’t send the platform bouncing between sites