Observability Quiz

Metrics, logging, and tracing. Figuring out what a production system is actually doing.

This category currently has 100 questions in the SERVBG quiz bank. Below are a few sample questions: the full interactive quiz shuffles through the whole set with instant scoring.

Sample questions

In Prometheus, what is the primary purpose of a recording rule?

To aggregate metrics from multiple Prometheus federation endpoints into a single namespace
To define alerting thresholds that trigger when a metric crosses a boundary
To relabel incoming metrics before they are stored in the TSDB
To precompute expensive or frequently evaluated PromQL expressions and save results as new time series
To configure scrape intervals on a per-target basis

Prometheus 2.50 introduced native histograms (also called sparse histograms). Which statement accurately describes a key behavioral difference from classic histograms?

Native histograms are stored in a separate TSDB block type incompatible with PromQL range queries
Native histograms store bucket boundaries dynamically in an exponential schema, eliminating the need to pre-declare le buckets at instrumentation time
Native histograms require a minimum scrape interval of 30s because exponential bucket computation is CPU-intensive
Native histograms replace the _sum and _count suffix metrics with a single merged sample for lower cardinality
Native histograms only work with OpenTelemetry exporters and cannot be scraped by standard Prometheus

What does the PromQL function `rate()` compute, and why must its range vector be at least 2x the scrape interval?

rate() returns the ratio of two counters over the range; 2x scrape interval is a Prometheus storage alignment requirement
rate() sums all counter resets in the range and divides by sample count; 2x scrape interval prevents integer overflow
rate() computes cumulative sum over the window; the 2x rule prevents double-counting at window boundaries
rate() computes the instantaneous derivative of a gauge; 2x scrape interval avoids stale marker collisions in the TSDB
rate() computes the per-second average increase of a counter over the range; at least 2x scrape interval ensures at least two data points exist for a meaningful calculation

A PromQL expression uses `histogram_percentile(0.95, rate(http_request_duration_seconds_bucket[5m]))`. This expression returns an error. What is the correct function name?

histogram_percentile is valid PromQL; the error is caused by a missing le label in the bucket metric series
quantile_over_time applied to the _bucket series produces equivalent output to the intended expression
histogram_quantile is the correct function; histogram_percentile does not exist in PromQL
percentile_over_time is the correct alternative function for computing percentiles over histogram buckets
The expression is syntactically valid but semantically wrong; use histogram_quantile with a subquery instead

In VictoriaMetrics, what is the primary architectural difference from vanilla Prometheus that enables horizontal scalability?

VictoriaMetrics uses consistent hashing at the scrape layer to distribute targets across Prometheus replicas automatically
VictoriaMetrics replaces the TSDB with a columnar Parquet store, requiring query translation from PromQL to SQL
VictoriaMetrics relies on external ZooKeeper coordination to shard metrics across storage nodes
VictoriaMetrics cluster mode splits ingestion (vminsert), storage (vmstorage), and query (vmselect) into independently scalable components
VictoriaMetrics uses a WAL-free write path that buffers all ingested data in memory until compaction

Start the full quiz ›