Support item count and operation name in azmonexporter (#2443)

* Support item count and operation name in azmonexporter

* Introduce AiSemanticAttributes

* Fix

* Comment

* Fix

* Fix

* Spotless

Author: trask

agent/agent-tooling/src/main/java/com/microsoft/applicationinsights/agent/internal/classicsdk/BytecodeUtilImpl.java

@@ -24,7 +24,9 @@
 import static java.util.concurrent.TimeUnit.MILLISECONDS;
 import static java.util.concurrent.TimeUnit.SECONDS;
 
-import com.azure.monitor.opentelemetry.exporter.implementation.AiOperationNameSpanProcessor;
+import com.azure.monitor.opentelemetry.exporter.implementation.AiSemanticAttributes;
+import com.azure.monitor.opentelemetry.exporter.implementation.OperationNames;
+import com.azure.monitor.opentelemetry.exporter.implementation.SamplingScoreGeneratorV2;
 import com.azure.monitor.opentelemetry.exporter.implementation.builders.AbstractTelemetryBuilder;
 import com.azure.monitor.opentelemetry.exporter.implementation.builders.AvailabilityTelemetryBuilder;
 import com.azure.monitor.opentelemetry.exporter.implementation.builders.EventTelemetryBuilder;
@@ -40,16 +42,12 @@
 import com.azure.monitor.opentelemetry.exporter.implementation.utils.FormattedDuration;
 import com.azure.monitor.opentelemetry.exporter.implementation.utils.FormattedTime;
 import com.azure.monitor.opentelemetry.exporter.implementation.utils.Strings;
-import com.azure.monitor.opentelemetry.exporter.implementation.utils.TelemetryUtil;
 import com.microsoft.applicationinsights.agent.bootstrap.BytecodeUtil.BytecodeUtilDelegate;
 import com.microsoft.applicationinsights.agent.internal.legacyheaders.AiLegacyPropagator;
-import com.microsoft.applicationinsights.agent.internal.sampling.SamplingScoreGeneratorV2;
 import com.microsoft.applicationinsights.agent.internal.statsbeat.FeatureStatsbeat;
 import com.microsoft.applicationinsights.agent.internal.telemetry.TelemetryClient;
 import io.opentelemetry.api.trace.Span;
 import io.opentelemetry.api.trace.SpanContext;
-import io.opentelemetry.context.Context;
-import io.opentelemetry.instrumentation.api.instrumenter.LocalRootSpan;
 import io.opentelemetry.sdk.trace.ReadableSpan;
 import java.net.URI;
 import java.net.URL;
@@ -472,78 +470,37 @@ public void logErrorOnce(Throwable t) {
   private static void track(
       AbstractTelemetryBuilder telemetryBuilder, Map<String, String> tags, boolean applySampling) {
 
-    String operationId = tags.get(ContextTagKeys.AI_OPERATION_ID.toString());
+    String existingOperationId = tags.get(ContextTagKeys.AI_OPERATION_ID.toString());
 
-    SpanContext context = Span.current().getSpanContext();
-    if (context.isValid()) {
-      String operationParentId = tags.get(ContextTagKeys.AI_OPERATION_PARENT_ID.toString());
-      String operationName = tags.get(ContextTagKeys.AI_OPERATION_NAME.toString());
+    Span span = Span.current();
+    SpanContext spanContext = span.getSpanContext();
 
-      trackInsideValidSpanContext(
-          telemetryBuilder, operationId, operationParentId, operationName, context, applySampling);
-    } else {
-      trackAsStandalone(telemetryBuilder, operationId, applySampling);
-    }
+    boolean isPartOfTheCurrentTrace =
+        spanContext.isValid()
+            && (existingOperationId == null
+                || existingOperationId.equals(spanContext.getTraceId()));
 
-    if (featureStatsbeat != null) {
-      featureStatsbeat.track2xBridgeUsage();
-    }
-  }
-
-  private static void trackInsideValidSpanContext(
-      AbstractTelemetryBuilder telemetryBuilder,
-      @Nullable String operationId,
-      @Nullable String operationParentId,
-      @Nullable String operationName,
-      SpanContext spanContext,
-      boolean applySampling) {
-
-    if (operationId != null && !operationId.equals(spanContext.getTraceId())) {
-      trackAsStandalone(telemetryBuilder, operationId, applySampling);
+    if (isPartOfTheCurrentTrace && applySampling && !spanContext.isSampled()) {
+      // no need to do anything more, sampled out
       return;
     }
 
-    if (!spanContext.isSampled()) {
-      // sampled out
-      return;
+    if (isPartOfTheCurrentTrace) {
+      setOperationTagsFromTheCurrentSpan(
+          telemetryBuilder, tags, existingOperationId, spanContext, span);
     }
 
-    telemetryBuilder.addTag(ContextTagKeys.AI_OPERATION_ID.toString(), spanContext.getTraceId());
-
-    if (operationParentId == null) {
-      telemetryBuilder.addTag(
-          ContextTagKeys.AI_OPERATION_PARENT_ID.toString(), spanContext.getSpanId());
-    }
-
-    if (operationName == null) {
-      Span localRootSpan = LocalRootSpan.fromContextOrNull(Context.current());
-      if (localRootSpan instanceof ReadableSpan) {
-        telemetryBuilder.addTag(
-            ContextTagKeys.AI_OPERATION_NAME.toString(),
-            AiOperationNameSpanProcessor.getOperationName((ReadableSpan) localRootSpan));
-      }
-    }
-
-    if (applySampling) {
-      float samplingPercentage =
-          TelemetryUtil.getSamplingPercentage(
-              spanContext.getTraceState(), BytecodeUtilImpl.samplingPercentage, false);
-
-      if (samplingPercentage != 100) {
-        telemetryBuilder.setSampleRate(samplingPercentage);
+    if (isPartOfTheCurrentTrace && applySampling && span instanceof ReadableSpan) {
+      Long itemCount = ((ReadableSpan) span).getAttribute(AiSemanticAttributes.ITEM_COUNT);
+      if (itemCount != null && itemCount != 1) {
+        telemetryBuilder.setSampleRate(100.0f / itemCount);
       }
     }
-    // this is not null because sdk instrumentation is not added until TelemetryClient.setActive()
-    // is called
-    TelemetryClient.getActive().trackAsync(telemetryBuilder.build());
-  }
 
-  private static void trackAsStandalone(
-      AbstractTelemetryBuilder telemetryBuilder, String operationId, boolean applySampling) {
-    if (applySampling) {
-      // sampling is done using the configured sampling percentage
+    if (!isPartOfTheCurrentTrace && applySampling) {
+      // standalone sampling is done using the configured sampling percentage
       float samplingPercentage = BytecodeUtilImpl.samplingPercentage;
-      if (!sample(operationId, samplingPercentage)) {
+      if (!sample(existingOperationId, samplingPercentage)) {
         logger.debug("Item {} sampled out", telemetryBuilder.getClass().getSimpleName());
         // sampled out
         return;
@@ -558,6 +515,33 @@ private static void trackAsStandalone(
     // this is not null because sdk instrumentation is not added until TelemetryClient.setActive()
     // is called
     TelemetryClient.getActive().trackAsync(telemetryBuilder.build());
+
+    if (featureStatsbeat != null) {
+      featureStatsbeat.track2xBridgeUsage();
+    }
+  }
+
+  private static void setOperationTagsFromTheCurrentSpan(
+      AbstractTelemetryBuilder telemetryBuilder,
+      Map<String, String> tags,
+      String existingOperationId,
+      SpanContext spanContext,
+      Span span) {
+
+    if (existingOperationId == null) {
+      telemetryBuilder.addTag(ContextTagKeys.AI_OPERATION_ID.toString(), spanContext.getTraceId());
+    }
+    String existingOperationParentId = tags.get(ContextTagKeys.AI_OPERATION_PARENT_ID.toString());
+    if (existingOperationParentId == null) {
+      telemetryBuilder.addTag(
+          ContextTagKeys.AI_OPERATION_PARENT_ID.toString(), spanContext.getSpanId());
+    }
+    String existingOperationName = tags.get(ContextTagKeys.AI_OPERATION_NAME.toString());
+    if (existingOperationName == null && span instanceof ReadableSpan) {
+      telemetryBuilder.addTag(
+          ContextTagKeys.AI_OPERATION_NAME.toString(),
+          OperationNames.getOperationName((ReadableSpan) span));
+    }
   }
 
   private static boolean sample(String operationId, double samplingPercentage) {

agent/azure-monitor-exporter/src/main/java/com/azure/monitor/opentelemetry/exporter/implementation/utils/TelemetryUtil.java renamed to agent/agent-tooling/src/main/java/com/microsoft/applicationinsights/agent/internal/classicsdk/TelemetryUtil.java

@@ -19,28 +19,25 @@
  * DEALINGS IN THE SOFTWARE.
  */
 
-package com.azure.monitor.opentelemetry.exporter.implementation.utils;
+package com.microsoft.applicationinsights.agent.internal.classicsdk;
 
 import com.azure.monitor.opentelemetry.exporter.implementation.builders.ExceptionDetailBuilder;
 import com.azure.monitor.opentelemetry.exporter.implementation.builders.StackFrameBuilder;
-import io.opentelemetry.api.trace.TraceState;
-import io.opentelemetry.instrumentation.api.internal.cache.Cache;
+import com.azure.monitor.opentelemetry.exporter.implementation.utils.Strings;
 import java.util.ArrayList;
 import java.util.List;
-import java.util.concurrent.atomic.AtomicBoolean;
 import javax.annotation.Nullable;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
-// naming convention:
-// * MonitorDomain data
-// * TelemetryItem telemetry
-public class TelemetryUtil {
+class TelemetryUtil {
 
-  private static final int MAX_PARSED_STACK_LENGTH =
-      32768; // Breeze will reject parsedStack exceeding 65536 bytes. Each char is 2 bytes long.
+  private static final Logger logger = LoggerFactory.getLogger(TelemetryUtil.class);
+
+  // Breeze will reject parsedStack exceeding 65536 bytes. Each char is 2 bytes long.
+  private static final int MAX_PARSED_STACK_LENGTH = 32768;
 
-  public static List<ExceptionDetailBuilder> getExceptions(Throwable throwable) {
+  static List<ExceptionDetailBuilder> getExceptions(Throwable throwable) {
     List<ExceptionDetailBuilder> exceptions = new ArrayList<>();
     convertExceptionTree(throwable, null, exceptions, Integer.MAX_VALUE);
     return exceptions;
@@ -150,74 +147,5 @@ private static int getStackFrameLength(@Nullable String text) {
     return text == null ? 0 : text.length();
   }
 
-  public static final String SAMPLING_PERCENTAGE_TRACE_STATE = "ai-internal-sp";
-
-  private static final Cache<String, OptionalFloat> parsedSamplingPercentageCache =
-      Cache.bounded(100);
-
-  private static final AtomicBoolean alreadyLoggedSamplingPercentageMissing = new AtomicBoolean();
-  private static final AtomicBoolean alreadyLoggedSamplingPercentageParseError =
-      new AtomicBoolean();
-
-  private static final Logger logger = LoggerFactory.getLogger(TelemetryUtil.class);
-
-  public static float getSamplingPercentage(
-      TraceState traceState, float defaultValue, boolean warnOnMissing) {
-    String samplingPercentageStr = traceState.get(SAMPLING_PERCENTAGE_TRACE_STATE);
-    if (samplingPercentageStr == null) {
-      if (warnOnMissing && !alreadyLoggedSamplingPercentageMissing.getAndSet(true)) {
-        // sampler should have set the trace state
-        logger.warn("did not find sampling percentage in trace state: {}", traceState);
-      }
-      return defaultValue;
-    }
-    return parseSamplingPercentage(samplingPercentageStr).orElse(defaultValue);
-  }
-
-  private static OptionalFloat parseSamplingPercentage(String samplingPercentageStr) {
-    return parsedSamplingPercentageCache.computeIfAbsent(
-        samplingPercentageStr,
-        str -> {
-          try {
-            return OptionalFloat.of(Float.parseFloat(str));
-          } catch (NumberFormatException e) {
-            if (!alreadyLoggedSamplingPercentageParseError.getAndSet(true)) {
-              logger.warn("error parsing sampling percentage trace state: {}", str, e);
-            }
-            return OptionalFloat.empty();
-          }
-        });
-  }
-
-  private static class OptionalFloat {
-
-    private static final OptionalFloat EMPTY = new OptionalFloat();
-
-    private final boolean present;
-    private final float value;
-
-    private OptionalFloat() {
-      this.present = false;
-      this.value = Float.NaN;
-    }
-
-    private OptionalFloat(float value) {
-      this.present = true;
-      this.value = value;
-    }
-
-    public static OptionalFloat empty() {
-      return EMPTY;
-    }
-
-    public static OptionalFloat of(float value) {
-      return new OptionalFloat(value);
-    }
-
-    public float orElse(float other) {
-      return present ? value : other;
-    }
-  }
-
   private TelemetryUtil() {}
 }

agent/agent-tooling/src/main/java/com/microsoft/applicationinsights/agent/internal/configuration/Configuration.java

@@ -156,6 +156,28 @@ public static class Sampling {
 
   public static class SamplingPreview {
 
+    // this is not the default for now at least, because
+    //
+    // parent not-sampled -> child not-sampled (always, to avoid broken traces)
+    // parent sampled -> child will still sample at its desired rate
+    //                   note: this is just sample rate of child, not sample rate of parent times
+    //                         sample rate of child, as both are using the same trace-id hash
+    //
+    // ??? if child sample rate is higher than the parent sample rate
+    //     parent sampled --> child sampled
+    //     parent not-sampled --> child not-sampled
+    //     which means that child has same effective sample rate as parent, and so its item count
+    //           will be wrong
+    //
+    // AND SO: if want to use parent-based sampler, then need to propagate the sample rate,
+    //         otherwise can end up with incorrect math
+    //
+    // Another (lesser) reason is because .NET SDK always propagates trace flags "00" (not
+    // sampled)
+    //
+    // IMPORTANT if changing this default, we need to keep it at least on Azure Functions
+    public boolean parentBased;
+
     public List<SamplingOverride> overrides = new ArrayList<>();
   }
 
@@ -261,10 +283,8 @@ public static class PreviewConfiguration {
     // world,
     // so safer to only allow single interval for now
     public int metricIntervalSeconds = 60;
-    // ignoreRemoteParentNotSampled is sometimes needed because .NET SDK always propagates trace
-    // flags "00" (not sampled)
-    // in particular, it is always needed in Azure Functions worker
-    public boolean ignoreRemoteParentNotSampled = DiagnosticsHelper.rpIntegrationChar() == 'f';
+    // this is just here to detect if using this old setting in order to give a helpful message
+    @Deprecated public Boolean ignoreRemoteParentNotSampled;
     public boolean captureControllerSpans;
     // this is just here to detect if using this old setting in order to give a helpful message
     @Deprecated public boolean httpMethodInOperationName;

agent/agent-tooling/src/main/java/com/microsoft/applicationinsights/agent/internal/configuration/ConfigurationBuilder.java

@@ -139,6 +139,13 @@ public static Configuration create(Path agentJarPath, @Nullable RpConfiguration
           "\"httpMethodInOperationName\" is no longer in preview and it is now the"
               + " (one and only) default behavior");
     }
+    if (config.preview.ignoreRemoteParentNotSampled != null) {
+      configurationLogger.warn(
+          "\"ignoreRemoteParentNotSampled\" has been deprecated and"
+              + " \"ignoreRemoteParentNotSampled\": false has replaced with"
+              + " \"preview\": { \"sampling\": { \"parentBased\": true } } (while"
+              + " \"ignoreRemoteParentNotSampled\": true is just the default behavior)");
+    }
     if (config.preview.openTelemetryApiSupport) {
       configurationLogger.warn(
           "\"openTelemetryApiSupport\" is no longer in preview and it is now the"

agent/agent-tooling/src/main/java/com/microsoft/applicationinsights/agent/internal/init/AppIdSupplier.java

@@ -29,7 +29,7 @@
 import com.azure.core.http.HttpResponse;
 import com.azure.monitor.opentelemetry.exporter.implementation.configuration.ConnectionString;
 import com.azure.monitor.opentelemetry.exporter.implementation.logging.NetworkFriendlyExceptions;
-import com.azure.monitor.opentelemetry.exporter.implementation.logging.WarningLogger;
+import com.azure.monitor.opentelemetry.exporter.implementation.logging.OperationLogger;
 import com.azure.monitor.opentelemetry.exporter.implementation.utils.ThreadPoolUtils;
 import com.microsoft.applicationinsights.agent.bootstrap.AiAppId;
 import com.microsoft.applicationinsights.agent.internal.httpclient.LazyHttpClient;
@@ -53,8 +53,8 @@ public class AppIdSupplier implements AiAppId.Supplier {
       Executors.newSingleThreadScheduledExecutor(
           ThreadPoolUtils.createDaemonThreadFactory(AppIdSupplier.class));
 
-  private static final WarningLogger warningLogger =
-      new WarningLogger(GetAppIdTask.class, "Unable to retrieve appId");
+  private static final OperationLogger operationLogger =
+      new OperationLogger(GetAppIdTask.class, "Retrieving appId");
 
   // guarded by taskLock
   private GetAppIdTask task;
@@ -147,7 +147,7 @@ public void run() {
       } catch (RuntimeException ex) {
         if (!NetworkFriendlyExceptions.logSpecialOneTimeFriendlyException(
             ex, url.toString(), friendlyExceptionThrown, logger)) {
-          warningLogger.recordWarning("exception sending request to " + url, ex, APP_ID_ERROR);
+          operationLogger.recordFailure("exception sending request to " + url, ex, APP_ID_ERROR);
         }
         backOff();
         return;
@@ -161,7 +161,7 @@ public void run() {
       String body = response.getBodyAsString().block();
       int statusCode = response.getStatusCode();
       if (statusCode != 200) {
-        warningLogger.recordWarning(
+        operationLogger.recordFailure(
             "received " + statusCode + " from " + url + "\nfull response:\n" + body,
             null,
             APP_ID_ERROR);
@@ -171,12 +171,12 @@ public void run() {
 
       // check for case when breeze returns invalid value
       if (body == null || body.isEmpty()) {
-        warningLogger.recordWarning("received empty body from " + url, null, APP_ID_ERROR);
+        operationLogger.recordFailure("received empty body from " + url, null, APP_ID_ERROR);
         backOff();
         return;
       }
 
-      logger.debug("appId retrieved: {}", body);
+      operationLogger.recordSuccess();
       appId = body;
     }