Course Test

Please answer the questions below to complete the test.

Question 1

What is the primary purpose of a Meshery performance profile?

Select one answer

To define Kubernetes resource quotas for a namespace

To store a reusable, reproducible load test specification that generates comparable result history over time

To configure Meshery Operator reconciliation intervals

To export Prometheus metrics to Grafana

Question 2

Which flag must you add to mesheryctl perf apply to capture the full latency histogram as structured data for programmatic processing?

Type your answer below

Question 3

fortio is the recommended default load generator for most API performance profiles because:

Select one answer

It is the only generator that supports HTTP/2

It runs at fixed QPS with coordinated omission correction, produces clean JSON output, and supports multiple protocols

It is the fastest generator available and can sustain the highest request rates

It is required by Meshery Operator for all performance scans

Question 4

When should you update the stored performance baseline? Select all that apply.

Select all that apply

After a verified optimization that measurably improves the profiled metrics

When an SLO or traffic pattern is intentionally changed

Whenever the gate fails, to prevent the pipeline from blocking

When the infrastructure changes in a way that makes the old baseline irrelevant

Question 5

A coding agent analyzing a performance profile observes: p50 = 18 ms, p99 = 22 ms, error rate = 4.3%, CPU utilization = 35%. What is the most likely root cause?

Select one answer

CPU saturation causing high tail latency

Connection pool exhaustion causing timeouts

A circuit breaker tripping or upstream dependency failures

The load generator is producing too many concurrent connections

Question 6

Which practices reduce false positive regression failures caused by normal performance variance? Select all that apply.

Select all that apply

Require a metric to exceed the budget by more than 15% before failing the gate

Run the profile three times and use the median result for comparison

Always run with a single concurrent connection to eliminate variance

Include a warm-up phase before the measurement window to avoid cold-start latency