intermediate

Metrics

Time-series measurements for service health, resource usage, user-visible behavior, and business outcomes.

Metrics interviews test time-series thinking: RED for APIs, USE for resources, business counters for product health, and Prometheus-style labels without cardinality explosions.

Subtopics: RED metrics, USE metrics, business metrics, Prometheus basics.

On interviews: pick three metrics for a GraphQL BFF; explain histogram buckets for latency SLO; when business metrics disagree with green infra dashboards.

Common pitfalls: unbounded route labels; averages instead of percentiles; no business counters on revenue paths.

The trade-off is rich dimensional data versus storage, query cost, and alert noise.

Checklist:

RED on request paths; USE on pools and nodes.
Bounded label sets.
Business metrics on critical funnels.
PromQL or equivalent for SLO views.