AZ-104 Practice Question 35: Autoscale Rule Evaluation
Microsoft Azure Administrator Associate (AZ-104) · Test 4 · question 5 of 10
A Virtual Machine Scale Set has one autoscale profile (minimum 2, maximum 20) containing four rules:
- Scale out by 2 when average CPU > 75% - Scale out by 5 when the storage queue length > 1000 - Scale in by 1 when average CPU < 25% - Scale in by 1 when the storage queue length < 100
During a burst, both scale-out conditions are true at the same time. An hour later, average CPU has fallen to 18% but the queue length is sitting at 400. Cooldown periods have elapsed.
What does autoscale do?
Show answer and explanation
Answer:
Autoscale deliberately biases toward availability, and the two halves work differently. On the way out: "If more than one scale-out rule is triggered, autoscale scales to the highest specified capacity to ensure service availability", so 5, not 7 and not 2. On the way in: autoscale evaluates scale-in rules only if no scale-out rule fired, and "autoscale only scales in if all the scale-in rules are triggered." With the queue at 400, the queue scale-in rule is false, so nothing is removed even though CPU is low.
Option A gets both halves wrong. Simultaneous scale-out actions are not summed, and the scale-in would not happen with one rule unmet.
Option C inverts the scale-out logic. Autoscale takes the *highest* capacity when several scale-out rules fire, and conservatism (removing the fewest instances) applies only to scale-in.
Option D reverses the evaluation order. Scale-out rules are evaluated first, and "if no scale-out rules are triggered, autoscale evaluates the scale-in rules." Autoscale also checks whether a scale-in would immediately trigger a scale-out, the flapping guard, and defers the action if so.