Prometheus for metrics collection in Kubernetes - Practice Problems & Coding Challenges

Challenge - 5 Problems

💻 Command Output

intermediate

Prometheus Query Result Interpretation

You run the Prometheus query rate(http_requests_total[5m]) to monitor HTTP requests per second over the last 5 minutes. What type of output do you expect from this query?

A A single integer representing the total number of HTTP requests since the server started.

B A time series showing the per-second rate of HTTP requests averaged over the last 5 minutes.

C A list of all HTTP request logs collected in the last 5 minutes.

D A histogram showing the distribution of HTTP request sizes.

💡 Hint

Think about what the rate() function does in Prometheus queries.

The rate() function calculates the per-second average increase of a counter over the specified time window. Here, it returns a time series showing the average HTTP requests per second over the last 5 minutes.

❓ Configuration

intermediate

Prometheus Scrape Configuration

You want Prometheus to scrape metrics from a Kubernetes service named my-app in the default namespace every 15 seconds. Which scrape configuration snippet correctly achieves this?

A

scrape_configs:
  - job_name: 'my-app'
    file_sd_configs:
      - files: ['my-app.yaml']
    scrape_interval: 15s

B

scrape_configs:
  - job_name: 'my-app'
    static_configs:
      - targets: ['my-app.default.svc.cluster.local:9090']
    scrape_interval: 15s

C

scrape_configs:
  - job_name: 'my-app'
    kubernetes_sd_configs:
      - role: endpoints
    relabel_configs:
      - source_labels: [__meta_kubernetes_service_name]
        action: keep
        regex: my-app
    scrape_interval: 15s

D

scrape_configs:
  - job_name: 'my-app'
    kubernetes_sd_configs:
      - role: pod
    scrape_interval: 15s

💡 Hint

Use Kubernetes service discovery and filter by service name.

Option C uses Kubernetes service discovery with role: endpoints and filters targets by the service name my-app. This is the correct way to scrape metrics from a Kubernetes service.

🔀 Workflow

advanced

Prometheus Alerting Workflow

You want to create an alert that fires when the CPU usage of any pod exceeds 80% for 5 minutes. Which sequence of steps correctly describes the workflow to achieve this?

A

1. Create a Kubernetes pod annotation for CPU limit.
2. Restart Prometheus.
3. Configure Grafana dashboard.
4. Send email manually when CPU is high.

B

1. Configure Prometheus to scrape node metrics.
2. Write a query for disk usage.
3. Restart Kubernetes cluster.
4. Configure Alertmanager to silence alerts.

C

1. Write a Prometheus query to list pods.
2. Use kubectl to scale pods.
3. Restart Alertmanager.
4. Configure Prometheus scrape interval.

D

1. Write a Prometheus alerting rule with a query checking CPU usage > 80% for 5 minutes.
2. Add the rule to Prometheus alerting rules file.
3. Reload Prometheus configuration.
4. Configure Alertmanager to handle the alert notifications.

💡 Hint

Focus on alerting rules and notification setup.

To alert on CPU usage, you write an alerting rule in Prometheus, load it, and configure Alertmanager to send notifications. Other options do not follow the correct alerting workflow.

❓ Troubleshoot

advanced

Prometheus Metrics Missing from Target

You notice Prometheus is not scraping metrics from a pod exposing metrics on port 9100, but the pod is healthy and reachable. What is the most likely cause?

A The pod's metrics endpoint is not labeled or annotated correctly for Prometheus to discover it.

B Prometheus server is down and not running.

C The pod is running on a node without network connectivity.

D The pod's container image is outdated.

💡 Hint

Check how Prometheus discovers scrape targets in Kubernetes.

Prometheus relies on labels or annotations to discover pods for scraping. If missing, Prometheus won't scrape metrics even if the pod is healthy.

✅ Best Practice

expert

Efficient Prometheus Metrics Collection Strategy

Which approach is the best practice to minimize Prometheus load while collecting metrics from a large Kubernetes cluster?

A Use service discovery with relabeling to scrape only necessary pods and endpoints, and increase scrape intervals for less critical metrics.

B Scrape all pods every 5 seconds to ensure real-time data accuracy.

C Disable service discovery and manually list all pod IPs in static configs.

D Run multiple Prometheus servers scraping the same targets simultaneously without coordination.

💡 Hint

Think about selective scraping and scrape frequency.

Selective scraping using service discovery with relabeling and adjusting scrape intervals reduces load and improves performance. Other options cause unnecessary load or complexity.