PCA Promql — 40 Questions

Question 1

An operator needs to query the current, un-extrapolated rate of increase per second over the last 5 minutes for a counter metric named http_requests_total. Which function should they use?

Accepted Answer

irate(http_requests_total[5m]). The irate() function calculates the per-second rate of increase of a time series based on the last two data points in the specified range window, making it ideal for volatile, high-frequency counters.

Answer

increase(http_requests_total[5m])

Answer

rate(http_requests_total[5m])

Answer

delta(http_requests_total[5m])

Question 2

An engineer needs to determine the total absolute increase in disk space consumed over the last 3 hours using a gauge metric node_disk_bytes_used. Which function is appropriate?

Accepted Answer

delta(node_disk_bytes_used[3h]). The increase() function is designed for counters, but delta() calculates the difference between the first and last value of a range vector, making it suitable for gauges.

Answer

deriv(node_disk_bytes_used[3h])

Answer

increase(node_disk_bytes_used[3h])

Answer

rate(node_disk_bytes_used[3h])

Question 3

Which label matcher operator is used for exact equality in PromQL selectors?

Accepted Answer

=. The = operator is the standard label matcher for exact string equality.

Answer

IS

Answer

:=

Answer

==

Question 4

You are joining an instant vector of application versions with an instant vector of metrics using a many-to-one relationship. The application version vector has more labels than the metric vector. Which keyword must be appended to the matching modifier to include labels from the 'many' side?

Accepted Answer

group_left. In many-to-one or one-to-many matches, group_left or group_right must be provided to specify which side supplies additional labels.

Answer

include_labels

Answer

keep_all

Answer

group_right

Question 5

An operator writes the query sum(rate(http_requests_total[5m])) without any aggregation clauses. What is the output format?

Accepted Answer

An instant vector containing a single element with no labels.. Aggregations without a by or without clause aggregate all time series into a single global vector with no labels.

Answer

A range vector spanning 5 minutes.

Answer

An instant vector preserving all original metric labels.

Answer

A scalar value representing the exact floating-point result.

Question 6

Which TWO of the following conditions will cause the rate() function to return NaN or empty results? (Choose TWO)

Accepted Answer

The time series has no samples within the specified range window.. Option B is correct because rate() operates over a range vector, and if the underlying time series has no samples inside that range window, the range vector is empty and rate() returns no result (an empty vector), not a numeric value. Option E is correct because rate() computes a per-second average increase from the first and last samples in the range; with fewer than two points there is no delta to divide by the window, so the function cannot produce a meaningful value and yields no result/NaN. Option A is not a cause: rate() is intended for counters, and feeding it a gauge may give misleading numbers, but it does not by itself force NaN or empty output. Option C is not a cause: rate() requires a range vector, and passing an instant vector is a query/type error rather than a rate() result of NaN or empty. Option D is not a cause: counter resets are handled by rate()'s extrapolation logic, which compensates for resets instead of returning NaN or empty.

Answer

The queried metric is a gauge rather than a counter, and samples do not follow monotonic rules.

Answer

The query is executed with an instant vector instead of a range vector.

Answer

The counter experienced a reset during the window.

Question 7

An operator writes a query to calculate error percentage: sum(rate(errors[5m])) / sum(rate(total[5m])) * 100. What happens if the denominator evaluates to zero or has no matching time series?

Accepted Answer

The query returns an empty vector (no data points).. In PromQL, when the denominator of a division has no matching series or evaluates to zero, the result is an empty vector rather than an error or NaN. Prometheus's vector matching semantics mean that if one side of the binary operation has no elements, there is nothing to pair with, so no output series are produced. This is why error-rate queries can silently return no data during periods with zero traffic.

Answer

Prometheus throws a fatal query evaluation runtime exception.

Answer

The query returns NaN for all series.

Answer

The query automatically returns infinity.

Question 8

You are joining an instant vector of application metrics with a static configuration instant vector containing target metadata using 'on(instance) group_right'. What does the 'group_right' modifier achieve?

Accepted Answer

It allows the right-hand side to provide more time series than the left-hand side, keeping all right-hand side series and including matched left-hand side data.. In a many-to-one or one-to-many match, group_right causes the metric on the right-hand side to be included for every matching left-hand side, keeping all labels from the right side and supplementing them.

Answer

It reverses the vector matching direction to evaluate from right to left.

Answer

It drops all labels from the left-hand side vector.

Answer

It ensures that the output contains only time series found exclusively on the right side.

Question 9

You need to calculate the 99th percentile of request latency across all instances using a Prometheus histogram metric 'http_request_duration_seconds_bucket'. Which function and setup is correct?

Accepted Answer

histogram_quantile(0.99, rate(http_request_duration_seconds_bucket[5m])). The histogram_quantile() function calculates the $\phi$-quantile from the buckets of a histogram metric. It requires the bucket metric and the quantile as arguments.

Answer

histogram_quantile(http_request_duration_seconds_bucket[5m], 0.99)

Answer

quantile_over_time(0.99, http_request_duration_seconds_bucket[5m])

Answer

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

Question 10

You want to aggregate total memory usage by 'job' using the 'node_memory_Usage_bytes' metric, while keeping the 'instance' label in the output. Which aggregation clause should you use?

Accepted Answer

sum(node_memory_Usage_bytes) by (job, instance). The 'without' clause removes listed labels from the result vector while keeping all other unlisted labels (such as instance). Alternatively, 'by' keeps only the specified labels.

Answer

sum(node_memory_Usage_bytes) drop (instance)

Answer

sum(node_memory_Usage_bytes) by (job)

Answer

sum(node_memory_Usage_bytes) without (instance)

Question 11

Which of the following selectors matches metric names that start with 'http_'?

Accepted Answer

{__name__=~"http_.*"}. The regex matcher '=~' is used with regular expressions, and '^' anchors the match to the beginning of the string.

Answer

http_*(5m)

Answer

{metric=~"http_.*"}

Answer

{__name__="http_*"}

Question 12

You need to calculate the ratio of successful HTTP requests (status 200) to total HTTP requests. Both metrics are counters named 'http_requests_total'. How do you construct this query to ensure proper label matching between status='200' and total requests?

Accepted Answer

http_requests_total{status="200"} / ignoring(status) http_requests_total. To divide two vectors with different label sets (e.g., status='200' vs all statuses), you must use binary matching modifiers like 'ignoring(status)' or 'on'.

Answer

rate(http_requests_total{status="200"}[5m]) / rate(http_requests_total[5m]) ignoring(status)

Answer

sum(http_requests_total{status="200"}) / sum(http_requests_total)

Answer

http_requests_total{status="200"} / http_requests_total

Question 13

Which of the following expressions correctly retrieves an instant vector of all HTTP requests with a 200 status code?

Accepted Answer

http_requests_total{status='200'}. Instant vectors return the most recent sample for each time series matching the selector.

Answer

rate(http_requests_total{status='200'})

Answer

http_requests_total{status='200'}[5m]

Answer

http_requests_total{status='200'} offset 5m

Question 14

When calculating the 95th percentile from a histogram, why is histogram_quantile(0.95, sum(rate(http_duration_seconds_bucket[5m])) by (le)) the recommended pattern?

Accepted Answer

Aggregating by 'le' keeps the bucket boundaries intact for the quantile calculation.. Aggregation must occur before the quantile calculation to ensure all buckets are summed across labels.

Answer

The 0.95 argument requires the 'le' label to be removed.

Answer

The rate function cannot be used inside histogram_quantile.

Answer

It is faster than calculating the quantile per instance.

Question 15

You have two metrics: 'up' (a gauge) and 'process_cpu_seconds_total' (a counter). You want to filter 'process_cpu_seconds_total' by instances where 'up == 1'. Which join modifier is required?

Accepted Answer

process_cpu_seconds_total * on(instance) up. The 'on' modifier specifies which labels to match, and 'group_left' or 'group_right' is used for many-to-one joins.

Answer

process_cpu_seconds_total * ignoring(job) up

Answer

process_cpu_seconds_total * group_left up

Answer

process_cpu_seconds_total and on(instance) up

Question 16

Which operator performs a logical 'not' on a comparison?

Accepted Answer

unless. The 'unless' set operator is used to filter out series that match the right-hand side.

Answer

!

Answer

not

Answer

none

Question 17

Which THREE of the following are valid time-range units in PromQL?

Accepted Answer

m. In PromQL, duration values appended to selectors and range vectors must use one of the supported time units: ms (milliseconds), s (seconds), m (minutes), h (hours), d (days), w (weeks), and y (years). Option B (m) is correct because m denotes minutes in PromQL durations, such as rate(http_requests_total[5m]). Option D (h) is correct because h denotes hours, e.g., [2h] for a two-hour range. Option E (s) is correct because s denotes seconds, e.g., [30s]. Options A (ns) and C (ms) are not both valid: ns (nanoseconds) is not a supported PromQL time unit, and although ms is valid, it is not marked as correct here, so it does not belong among the three selected answers.

Answer

ns

Answer

ms

Question 18

Which THREE of the following are valid ways to modify a binary operation join?

Accepted Answer

group_right. In LINQ, a binary operation join can be modified with group_right, on, and ignoring. Option A, group_right, is correct because it turns the join into a group join, producing a sequence of left elements each paired with matching right elements. Option B, on, is correct because it specifies the key selector expressions that determine how elements from the two sequences are matched. Option E, ignoring, is correct because it specifies a key comparer, such as StringComparer.OrdinalIgnoreCase, to control equality during the join. Option C, merge, is not a valid join modifier in LINQ, and option D, include, is not a valid join modifier either.

Answer

merge

Answer

include

Question 19

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 20

Which selector syntax correctly matches all time series where the environment label does NOT equal production?

Accepted Answer

http_requests_total{environment!="production"}. The != operator is the standard negative equality matcher in PromQL label selectors.

Answer

http_requests_total{environment=~"^((?!production).)*$"}

Answer

http_requests_total{environment!~"production"}

Answer

http_requests_total{environment!^"production"}