The cluster was failing and no one knew why. External requests stalled. Metrics flatlined. The root cause wasn’t CPU or memory—it was a broken promise of consistency. The external load balancer was cycling numbers like a slot machine, and every reconnect meant a new target.
Stable numbers for an external load balancer sound trivial until you see what happens without them. Microservices lose track of each other. Stateful connections die mid-transaction. Debugging turns into archaeology. The fix is simple in theory but rare in practice: assign unchanging, predictable numbers to the load balancer so every service, script, and firewall knows where to find it—always.
An external load balancer with stable numbers gives you more than resiliency. It gives you confidence. Firewall rules stay correct. IP allowlists don’t drift. Dev, staging, and prod environments stay in sync without manual patches. Rolling updates happen without an after-hours “who broke the cluster?” call.
Many teams confuse DNS stability with numeric stability. They’re not the same. DNS records can change, expire, or be cached in odd ways. Stable external numbers are explicit and respected. Networking becomes deterministic. Failovers are clean. Multi-region patterns become safer and simpler.