Prometheus Certified Associate (PCA) — Questions and Answers

Question 1: Which Prometheus service discovery mechanism discovers targets using Kubernetes API watch calls?

http_sd_config
file_sd_config
dns_sd_config
kubernetes_sd_config (Correct answer)

Correct answer: kubernetes_sd_config

`kubernetes_sd_config` uses the Kubernetes API to discover pods, services, endpoints, ingresses, and nodes as scrape targets.

Question 2: Which HTTP endpoint should you check on a Prometheus server to verify that a specific target is being scraped successfully?

/api/v1/alerts
/api/v1/targets (Correct answer)
/metrics
/api/v1/rules

Correct answer: /api/v1/targets

/api/v1/targets returns the current state of all scrape targets including their health status and last scrape error.

Question 3: Which exporter is commonly used to expose PostgreSQL database metrics for Prometheus?

postgresql_exporter
db_exporter
sql_exporter
postgres_exporter (Correct answer)

Correct answer: postgres_exporter

The `postgres_exporter` (by Prometheus Community) connects to PostgreSQL and exposes connection counts, query stats, replication lag, and more.

Question 4: What does the Prometheus blackbox_exporter probe?

Internal application JVM metrics
External endpoint availability via HTTP, HTTPS, DNS, TCP, and ICMP (Correct answer)
Container runtime metrics
Database query latency only

Correct answer: External endpoint availability via HTTP, HTTPS, DNS, TCP, and ICMP

The blackbox_exporter performs synthetic probes over HTTP(S), DNS, TCP, and ICMP to test external endpoint reachability and response time.

Question 5: A Grafana dashboard shows gaps in a Prometheus graph every hour. The target scrape is healthy. What is the most likely cause?

The range query interval exceeds the scrape interval causing no-data gaps (Correct answer)
The metric has a stale marker inserted every hour
Prometheus is restarting every hour due to OOM
The dashboard refresh rate is set to 1 hour

Correct answer: The range query interval exceeds the scrape interval causing no-data gaps

If the step (resolution) of a range query is larger than the scrape interval, Prometheus may find no samples in some steps, resulting in visible gaps.

Question 6: When Prometheus is deployed in a multi-cluster environment, which open-source project extends it with global query view and long-term storage without modifying the Prometheus binary?

Grafana Mimir
VictoriaMetrics
Loki
Thanos (Correct answer)

Correct answer: Thanos

Thanos sidecars attach to each Prometheus instance, uploading blocks to object storage and exposing a Store API, enabling a global query layer across clusters.

Question 7: What is stale marker handling in the context of Prometheus recording rules?

When a source series disappears, Prometheus marks recorded outputs as stale so dashboards show gaps instead of stale values (Correct answer)
It is a mechanism that replaces missing data points with the last known good value
Stale markers are injected into WAL to signal block compaction boundaries
Recording rules inject NaN values when source metrics exceed retention

Correct answer: When a source series disappears, Prometheus marks recorded outputs as stale so dashboards show gaps instead of stale values

Prometheus propagates staleness from input series to recording rule outputs, causing downstream metrics to also show as stale/missing when the source goes away.

Question 8: A Prometheus alert has been in the `PENDING` state for longer than the configured `for` duration but has not transitioned to `FIRING`. What is the most likely cause?

The rule evaluation interval is longer than the for-duration
Alertmanager is not reachable
The alert condition became false before the for-duration elapsed (Correct answer)
The alerting rule file has a syntax error

Correct answer: The alert condition became false before the for-duration elapsed

If the alert condition resolves before the for-duration completes, the alert resets to inactive without ever firing.

Question 9: What does the `address` label represent in Prometheus target discovery?

The host:port used to scrape the target (Correct answer)
The remote write endpoint
The Alertmanager URL
The Prometheus server's own address

Correct answer: The host:port used to scrape the target

`__address__` holds the `host:port` that Prometheus uses to construct the scrape URL for a target.

Question 10: What Prometheus label controls which Alertmanager instance a particular alert is routed to?

__replica__
There is no per-alert routing to specific Alertmanagers (Correct answer)
__alertmanager__
alertmanager_url

Correct answer: There is no per-alert routing to specific Alertmanagers

Prometheus sends all alerts to all configured Alertmanager instances; per-alert Alertmanager selection is not a Prometheus feature — routing is handled inside Alertmanager.

Question 11: Which Prometheus metric type is best suited for measuring request latency distributions with configurable quantile calculations?

Histogram
Counter
Gauge
Summary (Correct answer)

Correct answer: Summary

Summary calculates streaming quantiles (e.g., p50, p99) client-side and exposes them directly, making it suited when you need accurate per-instance quantiles.

Question 12: How does the `offset` modifier affect a PromQL query like `http_requests_total offset 1h`?

It averages the metric over the past 1 hour
It shifts the query evaluation time forward by 1 hour into the future
It delays the alert firing by 1 hour
It retrieves the metric value as it was 1 hour ago (Correct answer)

Correct answer: It retrieves the metric value as it was 1 hour ago

The `offset` modifier shifts the query's evaluation point backward in time, so `offset 1h` returns the value the metric had 1 hour ago.

Question 13: Which Grafana panel type is most appropriate for displaying a single current metric value such as the number of active connections?

Time series
Histogram
Heatmap
Stat (Correct answer)

Correct answer: Stat

The Stat panel displays a single large numeric value with optional coloring based on thresholds, ideal for KPI-style current metrics.

Question 14: What does Prometheus do when it encounters a 'stale marker' for a time series?

It immediately deletes the series from storage
It marks the series as stale and stops returning it in queries after the staleness window (Correct answer)
It retains the last known value indefinitely
It replaces the value with 0 until a new sample arrives

Correct answer: It marks the series as stale and stops returning it in queries after the staleness window

Prometheus sends stale markers when a target disappears, causing the series to be marked stale and excluded from query results after the staleness period (default 5 minutes).

Question 15: In Alertmanager, what is the difference between `resolve_timeout` and the `for` clause in a Prometheus alerting rule?

for controls when an alert fires; resolve_timeout controls how long Alertmanager waits before marking it resolved (Correct answer)
for applies only to critical alerts; resolve_timeout applies to all severities
They are identical; one is an alias of the other
resolve_timeout is set in Prometheus; for is set in Alertmanager

Correct answer: for controls when an alert fires; resolve_timeout controls how long Alertmanager waits before marking it resolved

The for clause in Prometheus determines how long a condition must be true before alerting, while resolve_timeout in Alertmanager sets how long to wait after the condition clears before sending a resolution notification.

Question 16: What is a dashboard in Grafana?

A query tool
A tool for storing data
A configuration panel
A visual representation of metrics (Correct answer)

Correct answer: A visual representation of metrics

The core purpose of monitoring with Prometheus is to systematically collect quantitative data (metrics) about the behavior and health of systems and applications. By continuously gathering these metrics, Prometheus enables users to observe performance trends, identify bottlenecks, detect anomalies, and ensure the overall stability and efficiency of their infrastructure. This comprehensive data collection forms the foundation of effective system management.

Question 17: Which Prometheus web configuration file field specifies the path to the server's TLS certificate?

server_certificate
tls_cert
ssl_cert_path
cert_file (Correct answer)

Correct answer: cert_file

The cert_file field in the web configuration YAML points to the PEM-encoded TLS certificate file for the Prometheus server.

Question 18: What is the recommended approach for instrumenting a new Go application with Prometheus?

Use the Pushgateway for all application metrics
Import the prometheus/client_golang library and register metrics with the default registry (Correct answer)
Write metrics to a file for file_sd to pick up
Embed the statsd_exporter as a sidecar

Correct answer: Import the prometheus/client_golang library and register metrics with the default registry

The official `prometheus/client_golang` library provides Counter, Gauge, Histogram, and Summary types that auto-register with the default registry and expose via an HTTP handler.

Question 19: When visualizing metrics in Grafana, what does 'legend format' allow you to customize?

The color and line style of each time series
The Y-axis unit and scale for the panel
The tooltip content shown on hover
The display name of each series in the legend using label values (Correct answer)

Correct answer: The display name of each series in the legend using label values

Legend format lets you define a template using label names (e.g., `{{instance}} - {{job}}`) to create meaningful legend labels for each time series.

Question 20: Which exporter is the standard choice for exposing Linux host-level metrics such as CPU, memory, and disk?

host_exporter
node_exporter (Correct answer)
blackbox_exporter
process_exporter

Correct answer: node_exporter

The `node_exporter` exposes a comprehensive set of hardware and OS metrics from Linux/Unix hosts via `/proc` and `/sys`.

Question 21: What is 'relabeling' used for in Prometheus's scrape configuration?

Compressing WAL segments
Transforming or filtering labels on targets or scraped samples before storage (Correct answer)
Renaming TSDB block files on disk
Assigning alert severity levels

Correct answer: Transforming or filtering labels on targets or scraped samples before storage

Relabeling (relabel_configs and metric_relabel_configs) allows operators to add, rename, drop, or filter labels at scrape time before samples are stored.

Question 22: What is the correct PromQL syntax for a subquery that evaluates `rate(http_requests_total[5m])` over the last 1 hour at 1-minute resolution?

`rate(http_requests_total[5m:1m])[1h]`
`rate(http_requests_total[5m])[1h:1m]` (Correct answer)
`rate(http_requests_total[5m])[1h]`
`subquery(rate(http_requests_total[5m]), 1h, 1m)`

Correct answer: `rate(http_requests_total[5m])[1h:1m]`

Subquery syntax appends `[range:resolution]` to a range expression, so `[1h:1m]` evaluates the inner expression over 1 hour at 1-minute steps.

Question 23: Which flag controls how long Prometheus retains data when using time-based retention?

--storage.tsdb.retention.time (Correct answer)
--tsdb.retention-days
--storage.local.retention
--storage.tsdb.max-age

Correct answer: --storage.tsdb.retention.time

The --storage.tsdb.retention.time flag specifies the maximum age of data blocks before they are deleted.

Question 24: What is the purpose of `ca_file` in a Prometheus tls_config block?

It lists the cipher algorithms allowed during scraping
It specifies the CA certificate used to verify the target's TLS certificate (Correct answer)
It names the Kubernetes secret containing TLS credentials
It defines the certificate authority for Prometheus's own HTTPS server

Correct answer: It specifies the CA certificate used to verify the target's TLS certificate

ca_file provides the CA bundle that Prometheus uses to validate the server certificate presented by scrape targets.

Question 25: In Prometheus HTTP service discovery (`http_sd_config`), what must the HTTP endpoint return?

A YAML list of scrape_configs
A DNS zone file
A Prometheus exposition format response
A JSON array of target groups with labels (Correct answer)

Correct answer: A JSON array of target groups with labels

`http_sd_config` expects the endpoint to return a JSON array of target group objects, each with a `targets` list and optional `labels` map.

Question 26: What is the recommended naming convention for a recording rule that pre-computes `sum(rate(http_requests_total[5m])) by (job)`?

`http_requests_total_rate_5m_sum`
`recording_rule_http_requests_total`
`job:http_requests_total:rate5m` (Correct answer)
`http_requests_sum_by_job`

Correct answer: `job:http_requests_total:rate5m`

Prometheus recommends the naming pattern `level:metric:operations` for recording rules, where level is the aggregation label, metric is the base metric, and operations describe the transformations applied.

Question 27: Which exporter is the standard choice for exposing Linux/Unix host-level metrics (CPU, memory, disk) to Prometheus?

Pushgateway
Node Exporter (Correct answer)
cAdvisor
Blackbox Exporter

Correct answer: Node Exporter

Node Exporter is the official Prometheus exporter that reads /proc and /sys to expose hardware and OS metrics from Linux hosts.

Question 28: When troubleshooting a failing alerting rule, which Prometheus API endpoint provides the current evaluation state and any errors for alerting rules?

/api/v1/alerts
/api/v1/rules (Correct answer)
/api/v1/targets
/api/v1/query

Correct answer: /api/v1/rules

/api/v1/rules returns all loaded alerting and recording rules along with their current state, last evaluation time, and any evaluation errors.

Question 29: In a Prometheus rule file, which YAML key defines a recording rule's output metric name?

name
record (Correct answer)
metric
output

Correct answer: record

The `record` key in a rule group entry specifies the name of the new time series that the rule's expression will be written to.

Prometheus Certified Associate (PCA)

The PCA certifies foundational knowledge of Prometheus monitoring and observability, covering PromQL, Prometheus architecture, instrumentation, exporters, and alerting. It is offered by the Linux Foundation and CNCF.

Questions

45 min

Time Limit

75%

To Pass

Checking for saved progress...

Exam Rules

You can skip questions and return to them later
Flag questions for review before submitting
No feedback shown until you submit the entire exam
Unanswered questions count as wrong — answer everything
10 pretest questions are mixed in and don't affect your score
Timer auto-submits when time runs out
Your progress is auto-saved every 30 seconds