A metric endpoint is an API access point that exposes measurable operational data for collection. Prometheus calls the designated endpoint on each Scylla node, retrieves its metrics, and stores them in specialized time-series storage. section 11.1.1 section 11.1.2
The endpoint supports monitoring questions such as whether read latency has increased, how many reads are in progress, and whether compactions have risen. Prometheus collects these values at a configured interval, 20 seconds by default in the Scylla monitoring stack, so their change over time can be queried with PromQL. section 11.1.2 section 11.1.2
Reading an exported metric
An endpoint can expose a counter such as scylla_cache_partition_insertions, with separate values labeled by shard. Prometheus scrapes these values and makes them available for time-based queries and dashboards. section 11.1.2
