1. Which signal is best for cheap, fast alerting on error rates?
- Traces only
- Logs only
- Metrics
- Screenshots
Show answer
Answer: Metrics
Aggregated numbers are cheap to evaluate.
Lesson: Metrics, Logs and Traces
25 questions from the Prometheus + Grafana course, each with the correct answer and an explanation. Try to answer before you open the answer.
Answer: Metrics
Aggregated numbers are cheap to evaluate.
Lesson: Metrics, Logs and Traces
Answer: It pulls them by scraping HTTP endpoints
Pull model with service discovery.
Lesson: How Prometheus Works
Answer: The metric name together with its full set of label values
Each label combination is a series.
Lesson: The Data Model
Answer: Gauge
It goes up and down.
Lesson: Counters and Gauges
Answer: Histogram buckets can be aggregated across instances before computing percentiles
You cannot average percentiles.
Lesson: Histograms and Summaries
Answer: Rate, Errors, Duration
A starting set for every service.
Lesson: Instrumenting an Application
Answer: Whether the last scrape of a target succeeded (1) or failed (0)
A built-in health signal.
Lesson: Scrape Configuration
Answer: node_exporter
Host-level metrics.
Lesson: Exporters
Answer: After scraping, on the scraped series
relabel_configs run before scraping.
Answer: =~
!~ is the negated form.
Lesson: Selectors and Matchers
Answer: Counters only grow; rate shows the per-second change and handles resets
The speed is what matters.
Lesson: rate, irate and increase
Answer: Adds series together, keeping only the route label
by keeps the listed labels.
Lesson: Aggregation
Answer: le
le identifies the bucket boundary.
Answer: Many-to-one matching, where several left series match one right series
Needed for info-metric joins.
Answer: Expensive queries are precomputed, so dashboards and alerts are fast
Computed once, read many times.
Lesson: Recording Rules
Answer: Requires the condition to stay true for that long before firing
Filters brief blips.
Lesson: Alerting Rules
Answer: Suppresses some alerts while a related, more important alert is firing
Reduces noise during incidents.
Lesson: Alertmanager
Answer: Checkout error ratio above the SLO burn threshold
Page on symptoms users feel.
Lesson: Designing Good Alerts
Answer: Heatmap
Buckets over time.
Lesson: Data Sources and Panels
Answer: The list of job label values found on the up metric
Populates dropdown options.
Lesson: Variables and Dashboard Design
Answer: Loads data sources and dashboards from files at start-up
Configuration as code.
Lesson: Dashboards as Code
Answer: user_id
Unbounded values create unbounded series.
Lesson: Cardinality
Answer: Run two identical Prometheus servers and let Alertmanager deduplicate
Simple HA by duplication.
Lesson: Storage, Retention and Scaling
Answer: Saturation
Utilisation, Saturation, Errors.
Lesson: RED and USE Methods
Answer: Paging alerts have runbook links
Actionable, maintainable monitoring.
Lesson: A Monitoring Checklist