[io.opentelemetry] Add OTLP metrics and traces

[Ltty]

Description

This is a follow-up to #21127 (initial logs-only contribution), extending the OpenTelemetry service with a full three-signal OTLP/HTTP pipeline - logs (already in #21127), metrics, and traces.

Metrics

A Micrometer OtlpMeterRegistry (Micrometer 1.16.3) is attached to openHAB's CompositeMeterRegistry via MeterRegistryProvider. A MeterFilter scopes export to openHAB-own meters (openhab.*, jvm.*, executor.*, process.*, system.*, logback.*, http.*) to avoid leaking unrelated add-on metrics.

Configuration: metricsEnabled, metricsEndpoint, metricsInterval (ISO-8601 duration), and metricsAggregationTemporality (CUMULATIVE / DELTA - required for some backends such as Dynatrace).

Traces (event-bus spans)

A SdkTracerProvider with BatchSpanProcessor and OtlpHttpSpanExporter is wired to an OpenTelemetryEventListener that subscribes to ALL_EVENT_TYPES on the openHAB event bus. Each event produces a flat root span (event.type, event.topic, event.source attributes). Spans are root-level by design - the event bus is fire-and-forget with no propagated call context.

When an OTel Java agent is detected at JVM startup the bundle reuses GlobalOpenTelemetry instead of starting a competing SDK instance.

Configuration: tracesEnabled, tracesEndpoint, tracesSamplingRatio (0.0-1.0, clamped, NaN-safe).

Lifecycle hardening

• Bounded forceFlush (3 s) before sdk.close() on deactivate/reconfigure
• Header injection guard in parseOtlpHeaders() - rejects \n, \r, \0
• Own-bundle and OTel-exporter log entries suppressed in OpenTelemetryLogListener to prevent export-failure feedback loops
• Cleartext http:// warning on startup
• Non-positive metricsInterval clamped to default (PT60S); NaN tracesSamplingRatio falls back to 1.0

OSGi dependency notes

micrometer-registry-otlp imports io.opentelemetry.proto.*. We embed opentelemetry-proto:1.8.0-alpha and protobuf-java:4.32.0 privately, and export io.opentelemetry.proto.* with -noimport:=true so the micrometer bundle can wire to them. All other OTel SDK classes stay private. The resulting JAR is self-contained.

Shared resource / signal correlation

All three signals share a single otlpURL / otlpHeaders and a common OTel Resource (service.name=openHAB, service.namespace, service.version, stable service.instance.id, host.name, os.*) so logs, metrics, and traces correlate to a single service entity in the observability backend.

Tests

31 unit tests added covering: header parsing and injection guard, URL joining (including vendor base-path regression), meter-name filter, log severity mapping, own-logger suppression, event attribute mapping, null/blank event.source omission, sampling-ratio clamping, and secret masking in toString().

Testing

Gate build: BUILD SUCCESS - 31/31 tests, 0 failures.

The three signals have been verified end-to-end against a Dynatrace OTLP ingest with both an OTel Collector and direct-to-backend configurations.

[Copilot]

Pull request overview

This PR extends the org.openhab.io.opentelemetry add-on from a logs-only OTLP/HTTP exporter to a full three-signal OpenTelemetry pipeline by adding OTLP metrics export (via Micrometer’s OTLP registry) and traces export (event-bus spans via an EventSubscriber). It also updates configuration/UI strings and expands unit test coverage for the new behavior.

Changes:

• Add Micrometer OTLP push metrics pipeline wired into openHAB’s CompositeMeterRegistry, with meter-name prefix filtering.
• Add event-bus span emission via a dynamically registered EventSubscriber, with optional reuse of GlobalOpenTelemetry when an OTel Java agent is detected.
• Update configuration schema/i18n/docs and add/extend unit tests for URL joining, header parsing hardening, filtering, and listener behaviors.

Reviewed changes

Copilot reviewed 14 out of 14 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
bundles/org.openhab.io.opentelemetry/src/main/java/org/openhab/io/opentelemetry/internal/OpenTelemetryService.java	Adds metrics + traces initialization/teardown, shared resource attributes, header injection guard, and agent coexistence logic.
bundles/org.openhab.io.opentelemetry/src/main/java/org/openhab/io/opentelemetry/internal/OpenTelemetryConfiguration.java	Adds metrics/traces configuration fields and replaces URI.resolve() with a safer URL join for vendor base-path endpoints.
bundles/org.openhab.io.opentelemetry/src/main/java/org/openhab/io/opentelemetry/internal/OpenTelemetryEventListener.java	New event-bus subscriber that emits one root span per event with basic attributes.
bundles/org.openhab.io.opentelemetry/src/main/java/org/openhab/io/opentelemetry/internal/OpenTelemetryLogListener.java	Adds suppression for bundle/exporter logs to avoid export-failure feedback loops; improves null handling.
bundles/org.openhab.io.opentelemetry/src/main/resources/OH-INF/config/config.xml	Adds metrics/traces parameter groups and new configuration parameters/options.
bundles/org.openhab.io.opentelemetry/src/main/resources/OH-INF/i18n/io-opentelemetry.properties	Adds i18n strings for new metrics/traces configuration and updates add-on description.
bundles/org.openhab.io.opentelemetry/src/main/resources/OH-INF/addon/addon.xml	Updates add-on description to include logs, metrics, and traces.
bundles/org.openhab.io.opentelemetry/README.md	Documents new signals, configuration parameters, deployment patterns, and agent coexistence.
bundles/org.openhab.io.opentelemetry/doc/otel-collector.md	Updates collector setup documentation for logs/metrics/traces pipelines and modernizes example configs.
bundles/org.openhab.io.opentelemetry/doc/lgtm-stack.md	Updates LGTM stack doc to reflect multi-signal support and temporality guidance.
bundles/org.openhab.io.opentelemetry/pom.xml	Adds Micrometer OTLP registry dependency and embeds/export-wires OTel proto/protobuf for OSGi.
bundles/org.openhab.io.opentelemetry/src/test/java/org/openhab/io/opentelemetry/internal/OpenTelemetryServiceTest.java	Expands tests for header validation, URL joining, meter filtering, config masking, and defaults.
bundles/org.openhab.io.opentelemetry/src/test/java/org/openhab/io/opentelemetry/internal/OpenTelemetryLogListenerTest.java	Adds test for suppression of bundle/exporter logger names.
bundles/org.openhab.io.opentelemetry/src/test/java/org/openhab/io/opentelemetry/internal/OpenTelemetryEventListenerTest.java	New tests validating event-bus subscription, span creation, attribute mapping, and exception safety.

Comments suppressed due to low confidence (2)

bundles/org.openhab.io.opentelemetry/README.md:130

• The README lists metricsEnabled defaulting to true, but the actual defaults in the add-on config (config.xml) and OpenTelemetryConfiguration are false. The documentation table should reflect the real default.

| `metricsEnabled` | Enable exporting openHAB metrics to the OTLP endpoint | `true` |

bundles/org.openhab.io.opentelemetry/README.md:139

• The README lists tracesEnabled defaulting to true, but the actual defaults in the add-on config (config.xml) and OpenTelemetryConfiguration are false. Please align the documentation default with the shipped configuration.

| `tracesEnabled` | Enable exporting event-bus spans to the OTLP endpoint | `true` |

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[wborn]

Thanks for extending the OpenTelemetry add-on with metrics and event-bus traces. The implementation is well documented, includes useful configuration validation, and adds good unit-test coverage. The previously reported issues also appear to have been addressed.

I found three remaining correctness issues that should be fixed before merging:

1. OpenTelemetry SDK loggers are no longer fully suppressed, so SDK-side exporter failures can still feed back into the log export pipeline.
2. Trace endpoint validation performed by OtlpHttpSpanExporterBuilder.setEndpoint() can escape the current error handling and fail component activation or reconfiguration.
3. The metrics filter infers ownership from broad name prefixes, even though openHAB core already identifies its meters using the openhab_core_metric=true tag.

As a non-blocking cleanup, the Micrometer OTLP registry should preferably use the same Micrometer version as openHAB core rather than pinning 1.16.3 while core uses 1.17.0.

[florian-h05]

Thanks for your PR, Florian!
I've added a few first comments -- README and tests haven't been reviewed yet.

Note: Please don't mark comments as resolved. I'll do that.

[florian-h05]

Ideally, make a small core PR to make them available publicly.
Ff the const is changed in one place, it will likely be forgotten to be changed in its duplicates.

[Ltty]

Agreed on the drift risk. However, I'd rather not block this one on that. I'll file the follow-up core PR and switch to the import once it lands. Let me know if you'd prefer it exposed via org.openhab.core.io.monitor directly or a new accessor method.

[florian-h05]

Agreed on not blocking this one here.
Add a new class to org.openhab.core.io.monitor (not the internal sub-package) holding the io.micrometer.core.instrument.Tag as a public static constant.

[Ltty]

Understood.

For now I extended the comment here so the duplicated literal is clearly marked as temporary.

[florian-h05]

Some final human review comments.
Will probably run an AI review on it as well.

[florian-h05]

Agreed on not blocking this one here.
Add a new class to org.openhab.core.io.monitor (not the internal sub-package) holding the io.micrometer.core.instrument.Tag as a public static constant.

[florian-h05]

Two minor comments, other then this LGTM.

[florian-h05]

LGTM, thanks!

@wborn FYI, I've run your AI review with Antigravity, it didn't find any thing besides what what I already commented on and #21219 (comment), which also has been addressed by now.

[wborn]

Thanks for the updates. The previously reported correctness issues are fixed in the current revision. There are two remaining issues that should be addressed before merge: the documented trace sampling configuration is ignored when event-bus spans are routed through the OTel Java agent, and the newly added third-party dependencies are not yet reflected in NOTICE.

This review was AI-assisted.

[wborn]

The findings from the previous review are addressed in the current revision: Java-agent sampling behavior is now documented and the added runtime dependencies are covered by NOTICE. The complete updated change was reviewed again and CI is successful.

One remaining non-blocking Java-agent detection edge case is noted inline.

This review was AI-assisted.

[lsiepel]

Besides the review, there is a DCO issue:
Commit sha: 63a5a3f, Author: Florian Lettner, Committer: Florian Lettner; The sign-off is missing.