Struct GatewayMetrics 

pub struct GatewayMetrics {
    pub registry: Registry,
    pub request_count: IntCounterVec,
    pub request_duration: HistogramVec,
    pub request_errors: IntCounterVec,
    pub node_execution_count: IntCounterVec,
    pub node_execution_duration: HistogramVec,
    pub node_errors: IntCounterVec,
    pub consumer_requests: IntCounterVec,
    pub counter_store_errors: IntCounterVec,
    pub cache_events: IntCounterVec,
    pub session_store_errors: IntCounterVec,
}

Global metrics registry for the gateway.

Owns a dedicated Prometheus [Registry] with all collectors pre-registered. All collectors are internally thread-safe, so a single instance is shared (behind an Arc) between the data-plane and the Admin API.

Fields

registry: Registry

Registry holding every collector below; gathered by Self::render.

request_count: IntCounterVec

gateway_requests_total — requests by route, method, status.

request_duration: HistogramVec

gateway_request_duration_seconds — end-to-end request latency histogram per route (buckets 1ms to 5s).

request_errors: IntCounterVec

gateway_request_errors_total — one increment per error record in Context.errors at the end of the request, by route and error_code.

This counts genuine node failures only. Deliberate outcome exits (denied, limited, broken, abort, redirect, preflight, routed, hit) append no error record and therefore never increment it — track those through gateway_requests_total’s status label.

node_execution_count: IntCounterVec

gateway_node_executions_total — graph node executions by policy, node_id, node_type.

node_execution_duration: HistogramVec

gateway_node_duration_seconds — per-node execution latency histogram by policy and node_id (buckets 0.1ms to 500ms).

node_errors: IntCounterVec

gateway_node_errors_total — node failures by policy, node_id, error_code.

consumer_requests: IntCounterVec

gateway_consumer_requests_total — requests attributed to an authenticated consumer, by route (parity with APISIX’s prometheus per-consumer counter). Recorded by the prometheus node, which must be placed after the auth node that attaches the consumer.

counter_store_errors: IntCounterVec

Counter-store (stores:) backend errors, per named store.

cache_events: IntCounterVec

Response-cache outcomes, per backend, store, and event.

store is the named stores: entry for policy: redis, and the empty string for policy: local (which has no store to name) — the same convention counter_store_errors/session_store_errors would use if they had a backend without one.

error is the one that matters: a cache degraded to always-miss keeps serving correct responses, just slower and with more upstream load, so it is invisible in every other signal.

session_store_errors: IntCounterVec

Session-store (stores:) backend errors, per named store.

Implementations

impl GatewayMetrics

pub fn new() -> Self

Creates a fresh registry with all gateway collectors registered.

Panics only if a collector cannot be built or registered, which is impossible with these fixed names/labels — treated as a programmer error at startup.

pub fn render(&self) -> String

Renders all registered metrics in the Prometheus text exposition format, as served by the Admin API’s /metrics endpoint.