Autoscaling translates observed metrics into replica changes. It cannot fix a shared bottleneck or compensate for incorrect application capacity assumptions.
Before you start
You should understand Pods, Deployments and Services. Read desired configuration separately from observed cluster state. Use a development cluster when trying changes, and inspect events and status rather than assuming that an accepted manifest means the workload is ready to serve traffic.
The practical goal is to reason through this situation: A queue worker may scale from backlog rather than average CPU. Read the walkthrough first, then try the interview exercise before opening its answer. The important part is explaining the decision and its consequences, rather than remembering a definition alone.
Step-by-step walkthrough
Step 1: Identify limiting capacity
CPU, backlog and dependency saturation can indicate different needs.
Step 2: Choose a meaningful metric
Scale workers from useful demand under known per-worker throughput.
Step 3: Protect shared dependencies
More replicas must not exceed database or partner capacity.
Worked scenario
A queue worker may scale from backlog rather than average CPU.
A queue grows because its database is saturated. Adding workers can increase contention and make each task slower, worsening the backlog. If worker CPU is the bottleneck instead, added replicas may help. Measure completion rate, queue age and dependency utilization together before choosing the scaling signal.
Common mistake
Increasing replicas can overload a database already at capacity.
Verify the behavior
Increase demand and observe scaling, stabilization and dependency load.
Interview exercise
Choose a scaling signal.
Answer and reasoning
Measure the limiting resource and user outcome, configure stabilization and keep dependency capacity within a safe budget.
Continue learning
Compare the scenario with the Kubernetes interview questions and test your understanding with the Kubernetes MCQs. For terminology and implementation details, consult the reference material.