☁️ Free practice test question from: KCNA (KCNA)

HPA is designed for rapid response to metric changes (typically CPU/memory utilization or custom metrics). It operates on a relatively short polling interval (usually 15-30 seconds by default) and can react within seconds or minutes to traffic spikes. When traffic suddenly increases, pod CPU/memory utilization rises quickly, triggering HPA almost immediately.

VPA works on a fundamentally different timescale and philosophy:

• It analyzes resource usage patterns over longer periods (hours or days)

• Makes more gradual adjustments based on observed trends

• Typically has a recommendation cooldown period

• Requires pod restarts to apply new resource settings

VPA is designed for optimization based on established patterns, not immediate reaction to spikes.

First, it's important to understand that VPA doesn't completely replace manual resource configuration. VPA operates in one of three modes:

1. Recommendation mode (Off): Only provides suggestions without applying them

2. Initial mode: Only applies recommendations when pods are first created

3. Auto mode: Can update existing pods (requiring restarts)

Cluster Autoscaler

The Cluster Autoscaler only activates when:

1. HPA has already created additional pods

2. These new pods enter a "Pending" state because there aren't enough resources on existing nodes

3. Only then does the Cluster Autoscaler add more nodes