Model your cluster, kill a node, and see whether the survivors can actually carry the load — before reality runs the test for you.
| Name | Cores | RAM GB |
|---|
| Name | Cores | RAM GB | Node | Priority |
|---|
Failover here is a greedy fit: critical guests place first, RAM is the hard constraint, CPU allows 2× overcommit. Real HA needs shared or replicated storage too — a guest whose disk died with its node isn't "migrating" anywhere. A Peira Labs field tool · peira.dev.