PCA Practice Exam 1 — 20 Questions

Question 1

What is the default retention period for Prometheus data?

Accepted Answer

15 days. The default retention for local Prometheus storage is 15 days.

Answer

1 year

Answer

30 days

Answer

7 days

Question 2

You have a Prometheus alerting rule that triggers too frequently during flapping states. Which feature should you use to prevent this without silencing the alert entirely?

Accepted Answer

The 'for' field. The 'for' field in a Prometheus alerting rule specifies how long the condition must be true before the alert fires. By setting a duration (e.g., '5m'), it prevents alerts from triggering during brief flapping states, as the condition must persist for that period. This reduces noise without silencing the alert entirely.

Answer

The 'resolve_timeout' parameter

Answer

The 'group_wait' parameter

Answer

The 'inhibit_rules' block

Question 3

Which TWO of the following are valid metric types natively supported by Prometheus? (Choose TWO)

Accepted Answer

Gauge. Prometheus natively supports four core metric types: Counter, Gauge, Histogram, and Summary. Option B (Gauge) is correct because a gauge represents a single numerical value that can arbitrarily go up and down, such as current memory usage or temperature. Option C (Counter) is correct because a counter is a cumulative metric that only increases (or resets to zero on restart), such as the total number of HTTP requests served. The other options are not valid Prometheus metric types: Matrix, Vector, and Scalar are PromQL data types (instant vector, range vector, scalar, string) used in query expressions, not metric types exposed by instrumented targets.

Answer

Matrix

Answer

Vector

Answer

Scalar

Question 4

Which TWO are common causes of high-cardinality in Kubernetes environments?

Accepted Answer

Using a unique transaction ID as a label.. High cardinality occurs when a metric label takes on a very large number of distinct values, multiplying the number of time series. Option B is correct because a unique transaction ID as a label creates a new series for every transaction, producing unbounded cardinality. Option C is correct because the pod name in a metric label changes with every pod restart or rollout, generating a new series per pod instance and causing cardinality to grow continuously. Options A, D, and E are not common causes of high cardinality: deployment names, cluster regions, and service names are relatively stable, low-cardinality dimensions with a limited set of values.

Answer

Using a deployment name label.

Answer

Using a cluster region label.

Answer

Using a service name label.

Question 5

Which TWO of the following are true about the Pushgateway?

Accepted Answer

Metrics must be deleted manually. Option D is correct because the Pushgateway never expires or garbage-collects pushed metrics on its own; pushed time series persist until they are explicitly removed, typically with an HTTP DELETE to a path such as /metrics/job/<job> (or via the pushgateway's admin API), so operators must delete stale metrics manually. Option E is correct because the Pushgateway exists specifically for short-lived, batch, or cron-style jobs that cannot be scraped reliably during their brief runtime; such jobs push their final metrics to the gateway, which then exposes them for Prometheus to scrape. Option A is wrong because the Pushgateway is not a drop-in replacement for standard scraping: it only holds pushed metrics and cannot discover or scrape targets, and it lacks the full semantics of a normal Prometheus scrape target. Option B is wrong because the Pushgateway is not a persistent metrics store; it keeps pushed metrics in memory and loses them on restart, with persistence being Prometheus's role. Option C is wrong because the Pushgateway does not automatically expire metrics; there is no built-in TTL, which is exactly why manual deletion is required.

Answer

It is a drop-in for standard scraping

Answer

It stores metrics persistently

Answer

It automatically expires metrics

Question 6

When using a Prometheus client library to instrument a Go application, which function call is typically used to register a new metric?

Accepted Answer

prometheus.MustRegister(). In Prometheus client libraries for Go, the prometheus.MustRegister() function is used to register a new metric with the default registry. It panics if the registration fails, ensuring that metrics are properly registered at startup. This is the typical way to make metrics available for scraping.

Answer

exporter.Start()

Answer

server.Listen()

Answer

client.InitMetric()

Question 7

How does PromQL handle operator precedence between arithmetic operators like multiplication (*) and addition (+)?

Accepted Answer

Multiplication has higher precedence than addition.. PromQL follows standard mathematical order of operations, where multiplication and division take precedence over addition and subtraction.

Answer

Operators are evaluated strictly from left to right regardless of type.

Answer

Operators are evaluated strictly from right to left.

Answer

Addition has higher precedence than multiplication.

Question 8

What is 'Monitoring Philosophy' regarding alerts?

Accepted Answer

Alert on symptoms, not causes.. The 'Monitoring Philosophy' in Google Cloud's operations suite (formerly Stackdriver) emphasizes alerting on symptoms—user-visible issues like high latency or error rates—rather than on causes such as CPU spikes or disk I/O. This approach reduces alert noise and focuses on what actually impacts users, aligning with SRE principles. By alerting on symptoms, you avoid false positives from normal fluctuations and ensure that alerts are actionable and tied to service level objectives (SLOs).

Answer

Alert on every CPU spike.

Answer

Only alert during business hours.

Answer

Alert on every error log.

Question 9

You need to prevent an alert from firing if a maintenance window is active. How should you approach this in Alertmanager?

Accepted Answer

Use a Silence. Silences are the standard way to prevent specific alerts from firing during a known maintenance period.

Answer

Delete the alerting rule

Answer

Modify the routing tree

Answer

Add a 'skip' label to the rule

Question 10

Which THREE of the following are valid label matchers?

Accepted Answer

!~. In Prometheus-style label matching, the valid matchers are = (equality), != (inequality), =~ (regex match), and !~ (negative regex match), so C (!~), D (=~), and E (!=) are all correct. C (!~) is right because it selects series whose label value does not match the given regular expression, e.g. job!~"api.*". D (=~) is right because it selects series whose label value matches the regular expression, e.g. job=~"api.*". E (!=) is right because it selects series whose label value differs from the given string, e.g. job!="api". Options A (>>) and B (==) are not valid matchers: >> is not a PromQL label operator at all, and == is a comparison operator used in expressions, not a label matcher.

Answer

>>

Answer

==

Question 11

What is the purpose of the 'up' metric?

Accepted Answer

To indicate if a target is reachable. The 'up' metric is 1 if the target was successfully scraped and 0 otherwise.

Answer

To show system uptime

Answer

To measure request latency

Answer

To track CPU usage

Question 12

You need to calculate the 99th percentile HTTP request latency from a Prometheus histogram named http_request_duration_seconds. Which function is correct?

Accepted Answer

histogram_quantile(0.99, rate(http_request_duration_seconds_bucket[5m])). The correct PromQL expression is histogram_quantile(0.99, rate(http_request_duration_seconds_bucket[5m])). This computes the 99th percentile by first calculating the per-second rate of increase of each bucket over a 5-minute window, then applying histogram_quantile to interpolate the quantile from the bucket boundaries. The rate function is essential to convert cumulative counters into a rate suitable for quantile estimation.

Answer

rate(histogram_quantile(0.99, http_request_duration_seconds_bucket)[5m])

Answer

quantile_over_time(0.99, http_request_duration_seconds[5m])

Answer

histogram_quantile(http_request_duration_seconds_bucket[5m], 0.99)

Question 13

Which function is used to convert a counter rate into a per-second rate over a time range?

Accepted Answer

rate(). rate() is the standard function for per-second rates of counters.

Answer

delta()

Answer

increase()

Answer

sum()

Question 14

You are troubleshooting high cardinality in your metrics. What is a common cause for this in a custom exporter?

Accepted Answer

Including dynamic values like IDs or timestamps in label values.. High cardinality occurs when a label value has an unbounded number of unique values (like a request ID or a timestamp), causing a massive explosion in the number of unique time series.

Answer

Using too many different metrics.

Answer

Scraping the exporter too frequently.

Answer

Using the wrong metric type for a simple count.

Question 15

Which THREE of the following are valid components within a Prometheus Alerting Rule file?

Accepted Answer

expr. Prometheus alerting rules contain a 'groups' array, which contains individual 'rules' (alerts or recordings), and each alert rule must have an 'alert' name and 'expr'.

Answer

receiver

Answer

routing_tree

Question 16

Which THREE of the following are valid aggregation operators in PromQL? (Choose THREE)

Accepted Answer

stddev. In PromQL, the aggregation operators are applied with syntax like ([parameter,] ) [without|by (...)], and the valid set includes stddev, sum, and topk. Option C (stddev) is correct because stddev is a built-in aggregation operator that computes the standard deviation across the aggregated time series. Option D (sum) is correct because sum is a core aggregation operator that adds the values of all matched series. Option E (topk) is correct because topk is a parameterized aggregation operator that returns the k series with the highest values. Options A (median) and B (average) are not valid PromQL aggregation operators; the average is expressed as sum(...) / count(...), and median is not provided as an aggregation operator in PromQL.

Answer

median

Answer

average

Question 17

A Prometheus administrator wants to query the current value of the metric 'http_requests_total'. Which of the following PromQL expressions should be used?

Accepted Answer

http_requests_total. The expression `http_requests_total` is a valid PromQL instant vector selector that returns the current value of the metric at the evaluation timestamp. It does not require a range or function because the question asks for the current value, not a rate or aggregation. This is the simplest and correct way to query a single metric's latest sample.

Answer

http_requests_total[5m]

Answer

rate(http_requests_total)

Answer

sum(http_requests_total) over (instance)

Question 18

How does Prometheus handle subqueries embedded within an instant query, such as rate(http_requests_total[5m])[30m:1m]?

Accepted Answer

It runs the inner instant query at multiple evaluation timestamps over the specified range at the given resolution, returning a range vector.. Prometheus evaluates subqueries by running the inner instant query at each step of the outer range, using the specified resolution (1m) and range (30m). This produces a range vector that can then be used by outer functions like rate(). The subquery syntax [30m:1m] means evaluate the inner expression over the last 30 minutes at 1-minute intervals.

Answer

It caches the inner query results in memory for 30 minutes.

Answer

It increases the global scrape interval to 1 minute for the next 30 minutes.

Answer

It converts the range vector into a scalar through statistical interpolation.

Question 19

Which TWO of the following statements regarding PromQL subqueries are correct? (Choose two)

Accepted Answer

Subqueries allow instant vector queries to be passed as range vector inputs to functions like max_over_time().. Option B is correct because a PromQL subquery wraps an instant vector expression and evaluates it over a time range, producing a range vector that can be fed into range-vector functions such as max_over_time(), avg_over_time(), or rate(). Option C is correct because the subquery syntax is exactly [:], for example max_over_time(rate(http_requests_total[5m])[1h:1m]). Option A is wrong because subqueries do not bypass the storage engine; they still query TSDB data and typically add overhead by evaluating the inner expression at each resolution step. Option D is wrong because subqueries operate on instant vector expressions (the inner expression must return an instant vector), not exclusively on range vectors. Option E is wrong because the resolution parameter is optional; when omitted, Prometheus uses the global evaluation_interval as the default step.

Answer

Subqueries execute faster than standard instant queries because they bypass the storage engine.

Answer

Subqueries can only be executed against range vectors, never instant vectors.

Answer

The resolution parameter in a subquery is mandatory.

Question 20

You need to send critical alerts to a specific Slack channel while warning alerts go to email. Where should this routing logic be defined?

Accepted Answer

In the Alertmanager 'route' configuration. The Alertmanager 'route' configuration is where routing logic is defined. It uses a tree of routes with matchers (e.g., severity='critical') to direct alerts to specific receivers (e.g., Slack for critical, email for warning). This is the core mechanism for conditional alert routing in Prometheus-based monitoring.

Answer

In the Alertmanager 'receivers' configuration

Answer

In the Grafana 'Alerting' UI

Answer

In the Prometheus 'alerting' rule file