High Availability

HostingOS runs an embedded Raft quorum cluster for high availability.

Key features

• One leader, never two (split-brain safety)
• Automatic failover with lease-gated service IPs
• Self-healing in under 60 seconds
• Watchdog self-fencing (off by default)
• Emergency recovery without a runbook

Tested scenarios

All tested live against a real cluster:

1. Standby power-off and rejoin
2. Active leader power-off and rejoin
3. Quorum loss and recovery
4. Split-brain (network partition)
5. Interrupted resync
6. IP takeover exclusivity
7. Leader restart