4.x Support OpenTelemetry metrics semantic conventions (server); allow selection of which endpoints collected for OTel, REST.request, and KPI metrics

[tjquinno]

Description

Resolves #11042
Resolves #11123

DRAFT - depends on #11044

Release Note

---

Selecting which requests are automatically measured

Helidon can automatically collect metrics for REST endpoint traffic:

• the automatic KPI metrics (simple counter of all requests or extended KPI metrics fi configured)
• the Helidon MP REST.request timers
• metrics prescribed by OpenTelemetry semantic conventions for HTTP server traffic (new feature - see below)

You can now control which REST endpoints are measured by adding configuration under server.features.observe.observer.metrics.auto to control which sockets are included in the measurements and which paths are enabled or disabled. This is very similar to the approach taken to configure selective tracing of REST endpoints.

The updated metrics doc pages describe this new feature in detail.

Collecting metrics from OpenTelemetry metrics semantic conventions

Users can now enable OpenTelemetry semantic conventions for metrics which measure incoming HTTP traffic by adding a dependency on the new io.helidon.webserver.observe:helidon-webserver-observe-telemetry-metrics component.

NOTE: Do not try to use OpenTelemetry semantic conventions for metrics with MicroProfile Metrics. The OpenTelemetry semantic conventions prescribe tag names which contain dots (".") which are illegal in MP Metrics tag names. MP Metrics rejects the attempt to register the OpenTelemetry semantic conventions metrics at runtime.

You can control which endpoints Helidon includes using the selection feature described just above.

The updated Helidon SE and MP metrics documentation pages describe both of these new features.

---

PR Overview

The existing metrics observer component now has a new SPI contract for providers of an HTTP filter that implements automatic HTTP metrics.

The metrics observer code finds all instances of the provider contract and invokes their filter method, passing the metrics observer configuration. By default, the provided filters are added for each socket; config can select which sockets should be so filtered.

The PR also contains a new component has an implementation of this provider contract.

This implementation registers and updates metrics in compliance with the OTel metrics HTTP semantic conventions. The code uses the existing Helidon neutral metrics API to register and update the meters. That means the timers are available via the metrics endpoint and programmatic access to the meter registry. Once #11043 is complete these metrics will be available via OTLP.

Existing code already implements some "legacy" HTTP traffic metrics (KPI metrics, either basic or extended). This PR also applies the new path selections added to the metrics observer config to those legacy metrics. If the user omits the auto

Documentation

Doc update included.

[tjquinno]

Replaced by #11168