You are designing autoscale rules for a production Azure Virtual Machine Scale Set (VMSS) … — Microsoft Azure Developer (AZ-204)

You are designing autoscale rules for a production Azure Virtual Machine Scale Set (VMSS) hosting a microservice API. The VMSS has a minimum instance count of $3$ and a maximum instance count of $10$ .

The current autoscale settings are configured as follows:
* Scale-out rule: When the CPU Percentage (aggregated as Average) is greater than $80\%$ for $10$ minutes, increase the instance count by $2$ .
* Scale-in rule: When the CPU Percentage (aggregated as Average) is less than $T\%$ for $10$ minutes, decrease the instance count by $2$ .

You need to configure the scale-in rule to prevent autoscale flapping (repeated scale-out and scale-in cycles) under constant workload conditions.

Which two changes should you implement? (Select two.)

Set the scale-in threshold $T$ to $45\%$ .Answer
Set the scale-in threshold $T$ to $55\%$ and change the scale-in action to decrease the instance count by $1$ .Answer
C
Set the scale-in threshold $T$ to $55\%$ .
D
Set the scale-in threshold $T$ to $65\%$ and change the scale-in action to decrease the instance count by $1$ .

Answer

To prevent autoscale flapping under constant workload, you should either set the scale-in threshold to 45% or set the scale-in threshold to 55% and change the scale-in action to decrease the instance count by 1.

To prevent autoscale flapping, the average CPU percentage after a scale-in event must remain below the scale-out threshold of 80%. When using the default configuration (decrement by 2), the most restrictive scale-in scenario occurs when scaling from 5 to 3 instances. A scale-in threshold of 45% ensures that the total workload is less than

5 \times 45\% = 225\%

, which translates to a post-scale-in average CPU of less than

75\%

on 3 instances. Alternatively, reducing the decrement to 1 instance allows a higher threshold of 55%. In this case, the most restrictive scenario is scaling from 4 to 3 instances. A threshold of 55% ensures the total workload is less than

4 \times 55\% = 220\%

, leading to a post-scale-in average CPU of less than

73.3\%

. Both options successfully prevent immediate scale-out.

Step-by-Step Solution

Analyze the scale-in condition when scaling from 5 to 3 instances with a decrement of 2.

The scale-in triggers when the average CPU is less than

T\%

, meaning the total workload is less than

5 \times T\%

. After scaling down to 3 instances, the new average CPU is

\frac{5 \times T\%}{3}

We must find a threshold

T

such that the new average CPU is strictly less than the scale-out threshold of 80% to avoid immediate scale-out.

Calculate the maximum safe threshold

T

for a decrement of 2.

\frac{5 \times T}{3} < 80 \implies 5T < 240 \implies T < 48\%

. Thus,

T = 45\%

is safe.

Setting

T

below 48% ensures that the post-scale-in CPU load remains below the scale-out trigger.

Analyze the scale-in condition when scaling from 4 to 3 instances with a decrement of 1.

The scale-in triggers when the average CPU is less than

T\%

, meaning the total workload is less than

4 \times T\%

. After scaling down to 3 instances, the new average CPU is

\frac{4 \times T\%}{3}

We need to verify if changing the decrement to 1 allows a higher threshold like 55%.

Calculate the maximum safe threshold

T

for a decrement of 1.

\frac{4 \times T}{3} < 80 \implies 4T < 240 \implies T < 60\%

. Thus,

T = 55\%

with a decrement of 1 is safe.

Setting

T

below 60% with a decrement of 1 ensures the post-scale-in CPU load remains below the scale-out trigger.

Key Concept

Autoscale flapping occurs when a scale-in action reduces capacity to a point where the remaining instances immediately exceed the scale-out threshold, causing an endless loop. To prevent this, the scale-in threshold and scale-in step size must be configured such that the workload at the trigger point, when distributed over the reduced instance count, does not exceed the scale-out threshold.

Estimated Time:3m 0s

Question

Answer

Step-by-Step Solution

Key Concept

Common Mistakes

Topics

More questions from this topic