Router: metricsInjection

Request Java downstream-latency injection into the metrics handler.

Type: Boolean. Default: false.

Meaning

The Java configuration describes injection of the downstream API’s elapsed response time into the metrics handler in the request/response chain. That timing includes downstream work and network latency, helping distinguish proxy overhead from backend time. It does not itself install a metrics handler, select an exporter, or enable a Prometheus endpoint.

metricsInjection: false

This is the declared default. The configuration description intends to disable the extra downstream metric injection.

metricsInjection: true
metricsName: router-response

Requests the extra Java downstream timing under router-response. An appropriate metrics handler must be available. Compare it with the gateway’s overall request timing to investigate where latency is spent; they measure different intervals.

Java implementation qualification: current RouterHandler.handleRequest() looks up an available metrics handler even when the flag is false and attaches timing metadata if one is found. Therefore the flag is not a reliable hard off-switch in that source revision.

Rust support: the configuration accepts the flag, but the current gateway does not use it to control its metrics recorder or reproduce Java’s named downstream injection. Rust request/stream metrics have their own runtime paths.

See Router configuration for Portal/values.yml syntax, runtime compatibility, and source references.