Explain the difference between desired, minimum, and maximum capacity.

0-2 years experience

1. 1You need to launch an Auto Scaling group for a web service that should normally run 4 instances but can scale down to 2 during low traffic. How would you set the desired, minimum, and maximum capacity?
2. 2If you set the desired capacity to 5, the minimum to 3, and the maximum to 4, what will happen when the group is created?
3. 3During a deployment you notice the Auto Scaling group has only 2 instances even though you set desired capacity to 3. What could cause that?

2-5 years experience

1. 1Your application experiences sudden spikes and you notice the Auto Scaling group is hitting its maximum capacity and requests are failing. Walk me through how you would troubleshoot and adjust the capacity settings.
2. 2Explain the trade‑offs of setting a high minimum capacity versus relying on scaling policies to add instances on demand.
3. 3During a scheduled scaling event you set a temporary desired capacity of 10, but after the event the group scales back down to 3 instead of the original 5. What might have gone wrong with the min/desired values?

5-8 years experience

1. 1Design a multi‑AZ Auto Scaling strategy for a critical service that must maintain at least 6 instances across three AZs, but you also want to limit total cost. How would you configure desired, minimum, and maximum capacity, and what additional mechanisms would you use?
2. 2Your team wants to use predictive scaling based on historical load. How does the concept of desired, min, max capacity interact with predictive scaling, and what edge cases could cause capacity drift?
3. 3If an instance fails health checks and the group replaces it, how does the minimum capacity affect the replacement timing, and what would you do to ensure high availability during a rolling upgrade?

8+ years experience

1. 1Across several services you’re consolidating Auto Scaling groups into a shared fleet to improve utilization. How would you decide on global min/max versus per‑service desired capacities, and what governance processes would you put in place?
2. 2A legacy system runs in a single Auto Scaling group with a static max capacity that is now a bottleneck. Describe a migration plan that re‑architects capacity controls while preserving SLA, considering cross‑team dependencies.
3. 3When designing a cloud‑native platform for multiple product teams, how would you expose capacity configuration (desired/min/max) as a self‑service API while preventing teams from mis‑configuring limits that could affect other workloads?