Add pending traces report in tracer flares

dd-trace-core/build.gradle

@@ -20,7 +20,7 @@ excludedClassesCoverage += [
   'datadog.trace.common.writer.RemoteMapper.NoopRemoteMapper',
   'datadog.trace.core.monitor.DDAgentStatsDConnection',
   'datadog.trace.core.monitor.LoggingStatsDClient',
-  'datadog.trace.core.PendingTraceBuffer.DelayingPendingTraceBuffer.FlushElement',
+  'datadog.trace.core.PendingTraceBuffer.DelayingPendingTraceBuffer.CommandElement',
   'datadog.trace.core.StatusLogger',
   // covered with CI Visibility smoke tests
   'datadog.trace.core.StreamingTraceCollector',

dd-trace-core/src/main/java/datadog/trace/common/writer/TraceDumpJsonExporter.java

@@ -0,0 +1,60 @@
+package datadog.trace.common.writer;
+
+import com.squareup.moshi.JsonAdapter;
+import com.squareup.moshi.Moshi;
+import com.squareup.moshi.Types;
+import datadog.trace.api.flare.TracerFlare;
+import datadog.trace.core.DDSpan;
+import java.io.IOException;
+import java.util.Collection;
+import java.util.List;
+import java.util.zip.ZipOutputStream;
+
+public class TraceDumpJsonExporter implements Writer {
+
+  private static final JsonAdapter<Collection<DDSpan>> TRACE_ADAPTER =
+      new Moshi.Builder()
+          .add(DDSpanJsonAdapter.buildFactory(false))
+          .build()
+          .adapter(Types.newParameterizedType(Collection.class, DDSpan.class));
+  private StringBuilder dumpText;
+  private ZipOutputStream zip;
+
+  public TraceDumpJsonExporter(ZipOutputStream zip) {
+    this.zip = zip;
+    dumpText = new StringBuilder();
+  }
+
+  public void write(final Collection<DDSpan> trace) {
+    dumpText.append(TRACE_ADAPTER.toJson(trace));
+    dumpText.append('\n');
+  }
+
+  @Override
+  public void write(List<DDSpan> trace) {
+    // Do nothing
+  }
+
+  @Override
+  public void start() {
+    // do nothing
+  }
+
+  @Override
+  public boolean flush() {
+    try {
+      TracerFlare.addText(zip, "pending_traces.txt", dumpText.toString());
+    } catch (IOException e) {
+      // do nothing
+    }
+    return true;
+  }
+
+  @Override
+  public void close() {
+    // do nothing
+  }
+
+  @Override
+  public void incrementDropCounts(int spanCount) {}
+}

dd-trace-core/src/main/java/datadog/trace/core/PendingTrace.java

@@ -7,6 +7,7 @@
 import datadog.trace.core.CoreTracer.ConfigSnapshot;
 import datadog.trace.core.monitor.HealthMetrics;
 import java.util.ArrayList;
+import java.util.Collection;
 import java.util.List;
 import java.util.concurrent.ConcurrentLinkedDeque;
 import java.util.concurrent.TimeUnit;
@@ -448,4 +449,8 @@ public static long getDurationNano(CoreSpan<?> span) {
     PendingTrace trace = (PendingTrace) traceCollector;
     return trace.getLastWriteTime() - span.getStartTime();
   }
+
+  Collection<DDSpan> getSpans() {
+    return spans;
+  }
 }

dd-trace-core/src/main/java/datadog/trace/core/PendingTraceBuffer.java

@@ -3,13 +3,22 @@
 import static datadog.trace.util.AgentThreadFactory.AgentThread.TRACE_MONITOR;
 import static datadog.trace.util.AgentThreadFactory.THREAD_JOIN_TIMOUT_MS;
 import static datadog.trace.util.AgentThreadFactory.newAgentThread;
+import static java.util.Comparator.comparingLong;
 
 import datadog.communication.ddagent.SharedCommunicationObjects;
 import datadog.trace.api.Config;
+import datadog.trace.api.flare.TracerFlare;
 import datadog.trace.api.time.TimeSource;
+import datadog.trace.common.writer.TraceDumpJsonExporter;
 import datadog.trace.core.monitor.HealthMetrics;
+import java.io.IOException;
+import java.util.ArrayList;
+import java.util.Comparator;
+import java.util.List;
 import java.util.concurrent.TimeUnit;
 import java.util.concurrent.atomic.AtomicInteger;
+import java.util.function.Predicate;
+import java.util.zip.ZipOutputStream;
 import org.jctools.queues.MessagePassingQueue;
 import org.jctools.queues.MpscBlockingConsumerArrayQueue;
 import org.slf4j.Logger;
@@ -47,13 +56,16 @@ private static class DelayingPendingTraceBuffer extends PendingTraceBuffer {
     private static final long FORCE_SEND_DELAY_MS = TimeUnit.SECONDS.toMillis(5);
     private static final long SEND_DELAY_NS = TimeUnit.MILLISECONDS.toNanos(500);
     private static final long SLEEP_TIME_MS = 100;
+    private static final CommandElement FLUSH_ELEMENT = new CommandElement();
+    private static final CommandElement DUMP_ELEMENT = new CommandElement();
 
     private final MpscBlockingConsumerArrayQueue<Element> queue;
     private final Thread worker;
     private final TimeSource timeSource;
 
     private volatile boolean closed = false;
     private final AtomicInteger flushCounter = new AtomicInteger(0);
+    private final AtomicInteger dumpCounter = new AtomicInteger(0);
 
     private final LongRunningTracesTracker runningTracesTracker;
 
@@ -78,6 +90,7 @@ public void enqueue(Element pendingTrace) {
 
     @Override
     public void start() {
+      TracerFlare.addReporter(new TracerDump(this));
       worker.start();
     }
 
@@ -108,10 +121,10 @@ public void flush() {
       if (worker.isAlive()) {
         int count = flushCounter.get();
         int loop = 1;
-        boolean signaled = queue.offer(FlushElement.FLUSH_ELEMENT);
+        boolean signaled = queue.offer(FLUSH_ELEMENT);
         while (!closed && !signaled) {
           yieldOrSleep(loop++);
-          signaled = queue.offer(FlushElement.FLUSH_ELEMENT);
+          signaled = queue.offer(FLUSH_ELEMENT);
         }
         int newCount = flushCounter.get();
         while (!closed && count >= newCount) {
@@ -130,9 +143,48 @@ public void accept(Element pendingTrace) {
       }
     }
 
-    private static final class FlushElement implements Element {
-      static FlushElement FLUSH_ELEMENT = new FlushElement();
+    private static final class DumpDrain
+        implements MessagePassingQueue.Consumer<Element>, MessagePassingQueue.Supplier<Element> {
+      private static final DumpDrain DUMP_DRAIN = new DumpDrain();
+      private static final int MAX_DUMPED_TRACES = 50;
+
+      private static final Comparator<Element> TRACE_BY_START_TIME =
+          comparingLong(trace -> trace.getRootSpan().getStartTime());
+      private static final Predicate<Element> NOT_PENDING_TRACE =
+          element -> !(element instanceof PendingTrace);
+
+      private volatile List<Element> DATA = new ArrayList<>();
+      private int index = 0;
 
+      @Override
+      public void accept(Element pendingTrace) {
+        DATA.add(pendingTrace);
+      }
+
+      @Override
+      public Element get() {
+        if (index < DATA.size()) {
+          return DATA.get(index++);
+        }
+        return null; // Should never reach here or else queue may break according to
+        // MessagePassingQueue docs
+      }
+
+      public List<Element> collectTraces() {
+        DATA.removeIf(NOT_PENDING_TRACE);
+        // Storing oldest traces first
+        DATA.sort(TRACE_BY_START_TIME.reversed());
+        List<Element> orderedTraces = DATA;
+        DATA = new ArrayList<>();
+        return orderedTraces;
+      }
+
+      public void clear() {
+        DATA.clear();
+      }
+    }
+
+    private static final class CommandElement implements Element {
       @Override
       public long oldestFinishedTime() {
         return 0;
@@ -180,13 +232,21 @@ public void run() {
               pendingTrace = queue.take(); // block until available;
             }
 
-            if (pendingTrace instanceof FlushElement) {
+            if (pendingTrace == FLUSH_ELEMENT) {
               // Since this is an MPSC queue, the drain needs to be called on the consumer thread
               queue.drain(WriteDrain.WRITE_DRAIN);
               flushCounter.incrementAndGet();
               continue;
             }
 
+            if (pendingTrace == DUMP_ELEMENT) {
+              queue.fill(
+                  DumpDrain.DUMP_DRAIN,
+                  queue.drain(DumpDrain.DUMP_DRAIN, DumpDrain.MAX_DUMPED_TRACES));
+              dumpCounter.incrementAndGet();
+              continue;
+            }
+
             // The element is no longer in the queue
             pendingTrace.setEnqueued(false);
 
@@ -208,7 +268,7 @@ public void run() {
               // Trace has been unmodified long enough, go ahead and write whatever is finished.
               pendingTrace.write();
             } else {
-              // Trace is too new.  Requeue it and sleep to avoid a hot loop.
+              // Trace is too new. Requeue it and sleep to avoid a hot loop.
               enqueue(pendingTrace);
               Thread.sleep(SLEEP_TIME_MS);
             }
@@ -277,4 +337,42 @@ public static PendingTraceBuffer discarding() {
   public abstract void flush();
 
   public abstract void enqueue(Element pendingTrace);
+
+  private static class TracerDump implements TracerFlare.Reporter {
+    private final DelayingPendingTraceBuffer buffer;
+
+    private TracerDump(DelayingPendingTraceBuffer buffer) {
+      this.buffer = buffer;
+    }
+
+    @Override
+    public void prepareForFlare() {
+      if (buffer.worker.isAlive()) {
+        int count = buffer.dumpCounter.get();
+        int loop = 1;
+        boolean signaled = buffer.queue.offer(DelayingPendingTraceBuffer.DUMP_ELEMENT);
+        while (!buffer.closed && !signaled) {
+          buffer.yieldOrSleep(loop++);
+          signaled = buffer.queue.offer(DelayingPendingTraceBuffer.DUMP_ELEMENT);
+        }
+        int newCount = buffer.dumpCounter.get();
+        while (!buffer.closed && count >= newCount) {
+          buffer.yieldOrSleep(loop++);
+          newCount = buffer.dumpCounter.get();
+        }
+      }
+    }
+
+    @Override
+    public void addReportToFlare(ZipOutputStream zip) throws IOException {
+      TraceDumpJsonExporter writer = new TraceDumpJsonExporter(zip);
+      for (Element e : DelayingPendingTraceBuffer.DumpDrain.DUMP_DRAIN.collectTraces()) {
+        if (e instanceof PendingTrace) {
+          PendingTrace trace = (PendingTrace) e;
+          writer.write(trace.getSpans());
+        }
+      }
+      writer.flush();
+    }
+  }
 }

dd-trace-core/src/test/groovy/datadog/trace/core/LongRunningTracesTrackerTest.groovy

@@ -168,7 +168,7 @@ class LongRunningTracesTrackerTest extends DDSpecification {
 
   PendingTrace newTraceToTrack() {
     PendingTrace trace = factory.create(DDTraceId.ONE)
-    PendingTraceBufferTest::newSpanOf(trace, PrioritySampling.SAMPLER_KEEP)
+    PendingTraceBufferTest::newSpanOf(trace, PrioritySampling.SAMPLER_KEEP, 0)
     return trace
   }