Prometheus Certified Associate (PCA) - Full Practice Exam 1

1. A checkout request calls five services. Logs show errors in several services, but the team needs the causal path and time spent in each call. Which telemetry is most useful?
2. A classic histogram has request_duration_seconds_sum and request_duration_seconds_count. What computes the average observed duration across the fleet over five minutes?
3. A classic histogram records request durations. Which expression estimates the fleet-wide 95th percentile while preserving bucket boundaries?
4. A classic latency histogram's boundaries are 0.1, 0.5, 1, and +Inf seconds. One 0.3-second request is observed. Which bucket counters increase?
5. A commonly used five-minute request rate query is expensive across many dashboards. Which Prometheus feature can precompute the derived series?
6. A counter increases from 100 to 130 during a range with regular scrapes and no resets. Which function estimates the total increase over the specified range rather than a per-second rate?
7. A dashboard shows no result for a queue metric after a target disappears. The team wants to detect lost monitoring coverage separately from an actual zero queue. What should it do?
8. A deployment metric is missing entirely for job="billing". Which expression can return an indicator when no matching series exists?
9. A developer names a cumulative duration metric request_time_milliseconds_total. Which naming change best follows Prometheus conventions when values are converted accordingly?
10. A fleet frequently adds and removes instances. Operators manually edit static scrape targets and miss newly started nodes. Which approach addresses discovery while retaining pull-based collection?
11. A left vector has one series per pod; a right vector has one metadata series per namespace. The query should attach a team label using namespace matching. Which modifier handles the many-pods-to-one-namespace relationship?
12. A library offers a CounterVec labeled by HTTP method and a normalized route. Which route value best avoids unbounded series growth?
13. A memory_bytes gauge has samples 100, 160, and 130 in the range. Which function returns 160?
14. A metric is named job_last_success_timestamp_seconds. Which expression gives seconds since the last successful job, assuming synchronized clocks?
15. A monitor checks a public endpoint from outside the service and records success and response time. What monitoring approach is this?
16. A new incident produces 80 alerts with the same cluster and service labels. The team wants a single initial grouped notification. Which configuration is relevant?
17. A nightly job finishes in two seconds and is not available when Prometheus next scrapes. The job records a service-level completion timestamp. Which approach fits this narrowly scoped use case?
18. A planned maintenance window should temporarily suppress notifications for matching instances without deleting alert rules. Which feature fits?
19. A query must select status codes 200 or 201 exactly, with no other values. Which selector fits?
20. A query should sum CPU consumption per team while dropping all other identifying dimensions. Which aggregation is appropriate?
21. A range contains a memory gauge falling from 900 to 600 bytes. Which function is appropriate for estimating its net change, including a negative result?
22. A regional Prometheus should expose selected aggregate series for a central Prometheus to scrape. Which feature is designed for this?
23. A request counter is scraped every 15 seconds. Which expression gives the average per-second increase over five minutes?
24. A scrape has interval 30s and timeout 10s. What happens when the target takes 12 seconds to respond fully?
25. A scrape job needs to collect /internal/metrics from servers on port 9105. Which configuration field changes the HTTP path?
26. A scrape succeeds at the HTTP layer, but the target returns malformed metric text. What value should the target's up metric have for that failed scrape?
27. A scrape target's process restarts and its request counter drops from 18,000 to 12. Which statement is correct?
28. A service objective requires 99.9% successful eligible requests per month. There are 2,000,000 eligible requests. How many failures consume the entire error budget?
29. A service reports 99.99% host uptime while users experience failed logins caused by an unavailable identity provider. What should the availability indicator measure?
30. A service tracks the number of requests currently being processed. The value rises and falls. Which metric type fits?
31. A target exposes useful metrics and one extremely costly debug metric family. The target must still be scraped, but those samples should be discarded before local ingestion. Which mechanism fits?
32. A target no longer appears in service discovery. What should a dashboard designer avoid assuming about a normal instant query for its old metric?
33. A team collects CPU utilization, error counts, logs, and traces but cannot answer whether releases harm checkout success. What should it define next?
34. A team configures two Prometheus servers to scrape the same services for availability. What is true of their local time-series storage?
35. A team exports the age of its last backup by continuously updating a gauge every second. Which alternative reduces update work while preserving the ability to calculate age?
36. A team is instrumenting function duration and must include failures in the measured distribution. What is the best placement of observation logic?
37. A team needs the percentage of HTTP requests returning 5xx across all instances. Both counters have identical scrape coverage. Which expression is appropriate?
38. A team needs to aggregate request latency distributions across instances and choose percentile queries after collection. Which instrumentation is most appropriate?
39. A team records each customer email as a metric label to troubleshoot individual purchases. Which redesign best preserves useful aggregate monitoring?
40. A team uses remote_write to send Prometheus samples to a remote system. Which component still normally collects the samples from applications?
41. A team wants to remove discovered targets whose environment label is test before any scrape is attempted. Which mechanism should it use?
42. A team's dashboard applies irate to a counter using the last two samples and sees brief spikes. It wants a steadier five-minute average for an alert. Which change is most appropriate?
43. A web service starts with no errors. Its error counter is created only when the first error occurs, causing missing series in dashboards. Which instrumentation improvement helps?
44. Alerts from one service should be delivered to its team, while unmatched alerts go to a default receiver. Where should this routing be configured?
45. An alert expression becomes true for 30 seconds during routine deployments. The team wants notification only after five continuous minutes of breach. Which rule setting fits?
46. An alert rule's expression is true but its for duration has not elapsed. What state is it in?
47. An alert should include a runbook link and a human-readable summary without changing its identifying label set whenever the text changes. Where should these be placed?
48. An API's average latency is stable, but premium users report very slow requests. Which signal best tests whether a small fraction of requests is slow?
49. An engineer edits prometheus.yml and wants to catch syntax and configuration errors before reloading the running server. Which command fits?
50. An engineer wants a series for every target, with 1 when up equals zero and 0 otherwise. Which expression fits?
51. An engineer wants the five instances with the highest current queue length, retaining their identifying labels. Which expression fits?
52. An existing database cannot expose Prometheus metrics directly. A supported exporter can query its operational statistics. What is the exporter's role?
53. An on-call team receives alerts for every short CPU spike, even when requests succeed. Which redesign best improves actionability?
54. An on-call team wants to find the exact error message for a failed transaction at 14:03. Which telemetry is the best starting point?
55. An organization promises customers a contractual availability level and separately tracks a stricter internal target. Which pairing is correct?
56. Hundreds of identical downstream alerts fire when the database becomes unavailable. The team wants notifications suppressed for downstream symptoms while the root-cause alert fires. Which mechanism fits?
57. The instant vector up contains values 1, 1, and 0 for three instances in job="api". What does sum(up{job="api"}) return?
58. Three instances expose a counter requests_total with a service label. Restarts occur independently. Which query best calculates request rate per service?
59. Two Prometheus replicas send the same alert to Alertmanager. Their replica label differs, preventing deduplication. Which change best supports deduplication while preserving other alert identity labels?
60. Two samples have the same metric name and labels except that one has region="east" and the other region="west". How are they represented?

Comments — 0