Vidocq Documentation
Version

Dirac is at 0.3.0. The official MicroProfile Metrics 5.1.1 TCK passes 127/127 (100 %) — see TCK.md. The public SPI may still evolve before 1.0.0. This page documents the mappings to prepare a progressive migration.

On the application side, migrating to Dirac is trivial: Dirac strictly consumes the standard annotations org.eclipse.microprofile.metrics.annotation.*. No proprietary import to replace. The deltas live in configuration, packaging and default exposition format.

From SmallRye Metrics

SmallRye Metrics is the reference implementation in Quarkus and some Jakarta EE servers. Direct migration — Dirac targets strict conformance to the same spec.

SmallRye Metrics Dirac Note

Annotations org.eclipse.microprofile.metrics.annotation.*

Same annotations

No application-code change.

/metrics endpoint

/metrics endpoint

Identical. OpenMetrics/JSON content negotiation identical.

mp.metrics.* (MP Config keys)

mp.metrics.distribution.* only

Percentiles and buckets: identical keys, read from system properties or microprofile-config.properties. Global tags (mp.metrics.tags, mp.metrics.appName): not implemented in Dirac.

Dependencies: Micrometer, HdrHistogram, Jackson…

None (hand-written formatters)

Smaller classpath footprint. Immediate jlink compatibility (M9).

JSON format via Jackson

JSON format via hand-written StringBuilder

No external JSON engine — check clients that rely on key ordering.

From Micrometer

Micrometer is the default instrumentation API for Spring Boot. Migrating to MP Metrics implies swapping the application-side API.

Micrometer Dirac (MP Metrics) Note

@Counted("metric") (io.micrometer.core.annotation)

@Counted(name = "metric") (org.eclipse.microprofile.metrics.annotation)

Import change. Same semantics.

Counter.builder("name").tags(…​).register(registry)

registry.counter(Metadata.builder().withName("name").build(), tags)

Different programmatic API, equivalent behaviour.

MeterRegistry (io.micrometer.core.instrument)

MetricRegistry (org.eclipse.microprofile.metrics)

Injection via @Inject + @RegistryScope qualifier.

Backend-specific exporters (Prometheus, Datadog, StatsD…)

Native OpenMetrics text + JSON

For other backends, plug a Prometheus scraper → destination exporter.

From Dropwizard Metrics

Dropwizard is the ancestor of much of the MP Metrics spec. Two-step migration: change the API, then expose.

Dropwizard Dirac (MP Metrics) Note

MetricRegistry.counter("name")

registry.counter(metadata, tags)

Tags concept absent in Dropwizard — switch to MP tags.

Timer.time(Runnable)

@Timed or Timer.time(Runnable)

Equivalent runtime API; prefer the annotation for CDI-intercepted code.

Histogram.update(value)

Histogram.update(value)

Identical. Percentiles configured via mp.metrics.distribution.percentiles.

Exposition via MetricsServlet + prometheus-dropwizard adapter

Native OpenMetrics + JSON exposition on /metrics

Drop the adapter and the servlet — the MP Metrics endpoint suffices.

Gotchas

  • No Meter, no ConcurrentGauge, no SimpleTimer — removed in MP Metrics 5.0. Replace with Counter + Timer as appropriate.

  • @Histogram (annotation) — absent from the microprofile-metrics-api:5.1.1 JAR used by Dirac. Declare histograms programmatically via MetricRegistry.histogram(…​).

  • No setAccessible(true) — if application code relied on non-public classes scanned by a permissive scanner, Dirac will not reach them. Make types public or add opens in the application module-info.

  • Fixed endpoint — /metrics is mandated by the spec. To publish at another path, add a JAX-RS alias on the application side.

  • Global tags not implemented — mp.metrics.tags and mp.metrics.appName are not read by Dirac, and no mp_scope label is emitted (the registry rejects mp_scope/mp_app as user tags). If the previous stack relied on global tags, declare the tags on each metric instead.

Porting checklist

  1. Declare dirac-core + dirac-cdi-vauban + dirac-rest instead of SmallRye / Micrometer / Dropwizard.

  2. Replace proprietary imports (io.micrometer., com.codahale., io.smallrye.metrics.) with org.eclipse.microprofile.metrics..

  3. Migrate proprietary properties to mp.metrics.* keys (Reference).

  4. Run ./mvnw test then ./run-official-tck-mp-metrics-5.1.sh all.

  5. Compare OpenMetrics output before/after — check canonical names, per-metric tags, units, the _total/_seconds sample suffixes and the final # EOF marker.

Further reading