The service should ___ a metric every time a request is handled.
To emit a metric means to produce and send a measurement from your code to a monitoring system. It is the standard observability collocation: "the service emits latency metrics". "Throw out", "spit", and "give" are not idiomatic. Emit pairs with metrics, events, and logs, and signals that the application actively reports telemetry.
2 / 5
Prometheus will ___ the /metrics endpoint every 15 seconds.
To scrape a target means to periodically pull its metrics from an exposed endpoint. It is the precise term in pull-based monitoring: "Prometheus scrapes the exporter". "Suck", "pull off", and "grab" are not the technical word. Scrape pairs with endpoint, target, and interval, and distinguishes pull-based systems from push-based ones.
3 / 5
The dashboard will ___ per-pod metrics into a single cluster view.
To aggregate metrics means to combine many data points (sum, average, percentile) into a higher-level value. It is the standard collocation: "aggregate by service", "aggregated latency". "Add up", "pile", and "collect over" are vaguer. Aggregate pairs with metrics, data, and logs, and is essential for turning raw telemetry into meaningful signals.
4 / 5
We should ___ on error rate so we get paged at 5%.
To set a threshold means to define the value at which an alert should trigger. It is the precise alerting collocation: "set a threshold on 95th-percentile latency". "Put a limit", "place a line", and "fix a level" are not idiomatic. Threshold pairs with set, breach, and exceed, and is central to defining alert conditions.
5 / 5
When latency crosses the limit, the rule will ___ an alert.
To fire an alert means for an alerting rule to trigger and notify on-call engineers. It is the standard collocation: "the alert fired at 2am". "Shoot", "blast", and "launch off" are not used. Fire pairs with alert, rule, and trigger, and describes the moment a monitored condition is breached.