“If NYC shuts down power how long can we run on diesel at the datacenter?”
That was the whole email.
Hurricane Irene was moving toward New York. Transit was stopping, low-lying areas were evacuating, and the city was preparing for conditions most of our normal plans did not cover.
The storm had already made staffing a problem. Nobody should be sent across the city during dangerous conditions simply because a runbook assumed a person could reach the facility.
For the website, the question was fuel.
A generator is reassuring until the utility stays off longer than the tank lasts. Then runtime depends on consumption, available diesel, delivery access, and whether anybody can safely reach the building.
Fuel contracts were only useful if a truck could arrive. Monitoring was only useful if somebody had a decision attached to the number.
The data center could have redundant power equipment and still depend on one city whose streets were about to become difficult.
This was why disaster planning had to use plain questions. “Do we have generators?” was too easy. “How many hours?” forced the next layer of the plan into view.
If the answer was not enough, we needed to reduce load, find more fuel, move service, or accept a different risk before the storm made the decision for us.
There was no perfect hurricane configuration available that afternoon. There was only time to confirm the real constraints and make the least bad plan.
I sent the question from my iPhone and waited for the number.
Outside, New York was getting ready to shut down. Inside the infrastructure plan, the clock would start when the power stopped.
How long on diesel?