Support OTLP runtime metrics with OTel-native naming (#11318)

Adding configs and metrics creation

Adding test file to check metrics are collected

Doing clean up after testing

Merge branch 'master' into maximo/otlp-runtime-metrics

Merge branch 'master' into maximo/otlp-runtime-metrics

move JvmOtlpRuntimeMetrics.java to agent-jmxfetch

prevent JMXFetch from emitting jvm metrics when otlp is enabled; migrate from depending on otel-shim to otel-bootstrap

update JMXFetch to only emit either OTLP or JMX runtime metrics

Merge branch 'master' into maximo/otlp-runtime-metrics

send otlp_jmx_config when otlp runtime metrics enabled

update test to assert on guarantees instead of dependent on GC collection

Merge branch 'master' into maximo/otlp-runtime-metrics

adding exception handling for callback

Merge remote-tracking branch 'origin/master' into maximo/otlp-runtime-metrics

Minor fixes to use correct storage for observable counters

Cleanup

Merge branch 'master' into maximo/otlp-runtime-metrics

Co-authored-by: mhlidd <matthew.li@datadoghq.com>
Co-authored-by: mcculls <stuart.mcculloch@datadoghq.com>
Co-authored-by: devflow.devflow-routing-intake <devflow.devflow-routing-intake@kubernetes.us1.ddbuild.io>

Author: 4 people

dd-java-agent/agent-jmxfetch/build.gradle

@@ -25,6 +25,11 @@ dependencies {
   api libs.slf4j
   api project(':internal-api')
   api project(':dd-java-agent:agent-bootstrap')
+
+  // JvmOtlpRuntimeMetrics registers JVM runtime instruments directly against the
+  // bootstrap-level OTel metric registry. otel-bootstrap vendors and repackages the
+  // OTel API at build time so this won't conflict with anything in the customer app.
+  compileOnly project(path: ':dd-java-agent:agent-otel:otel-bootstrap', configuration: 'shadow')
 }
 
 tasks.named("shadowJar", ShadowJar) {

dd-java-agent/agent-jmxfetch/src/main/java/datadog/trace/agent/jmxfetch/JMXFetch.java

@@ -9,6 +9,7 @@
 import datadog.metrics.api.statsd.StatsDClientManager;
 import datadog.trace.api.Config;
 import datadog.trace.api.GlobalTracer;
+import datadog.trace.api.InstrumenterConfig;
 import datadog.trace.api.flare.TracerFlare;
 import datadog.trace.api.telemetry.LogCollector;
 import de.thetaphi.forbiddenapis.SuppressForbidden;
@@ -32,6 +33,7 @@ public class JMXFetch {
   private static final Logger log = LoggerFactory.getLogger(JMXFetch.class);
 
   private static final String DEFAULT_CONFIG = "jmxfetch-config.yaml";
+  private static final String OTLP_JMX_CONFIG = "jmxfetch-config-no-jvm-defaults.yaml";
   private static final String WEBSPHERE_CONFIG = "jmxfetch-websphere-config.yaml";
 
   private static final int DELAY_BETWEEN_RUN_ATTEMPTS = 5000;
@@ -93,9 +95,22 @@ private static void run(final StatsDClientManager statsDClientManager, final Con
     final StatsDClient statsd = statsDClientManager.statsDClient(host, port, namedPipe, null, null);
     final AgentStatsdReporter reporter = new AgentStatsdReporter(statsd);
 
+    final boolean otlpRuntimeMetricsEnabled =
+        InstrumenterConfig.get().isMetricsOtelEnabled() && config.isMetricsOtlpExporterEnabled();
+
     TracerFlare.addReporter(reporter);
     final List<String> defaultConfigs = new ArrayList<>();
-    defaultConfigs.add(DEFAULT_CONFIG);
+    if (otlpRuntimeMetricsEnabled) {
+      // Register JVM runtime metric callbacks against the OtelMeterProvider so the OTLP
+      // exporter started by CoreTracer collects them. Started here so it rides the same
+      // delayed-start path as JMXFetch itself.
+      JvmOtlpRuntimeMetrics.start();
+      // When the OTLP exporter is collecting JVM runtime metrics, skip the default JMXFetch
+      // JVM config to avoid double-reporting.
+      defaultConfigs.add(OTLP_JMX_CONFIG);
+    } else {
+      defaultConfigs.add(DEFAULT_CONFIG);
+    }
     if (config.isJmxFetchIntegrationEnabled(Collections.singletonList("websphere"), false)) {
       defaultConfigs.add(WEBSPHERE_CONFIG);
     }