add Kafka sdk poll metrics

apache · Nov 10, 2024 · 3187c45 · 3187c45
1 parent 7ba2869
commit 3187c45
Show file tree

Hide file tree

Showing 6 changed files with 81 additions and 34 deletions.
diff --git a/...le-cloud-dataflow-java/src/main/java/org/apache/beam/runners/dataflow/DataflowRunner.java b/...le-cloud-dataflow-java/src/main/java/org/apache/beam/runners/dataflow/DataflowRunner.java
@@ -98,6 +98,7 @@
 import org.apache.beam.sdk.io.gcp.pubsub.PubsubUnboundedSource;
 import org.apache.beam.sdk.io.gcp.pubsublite.internal.SubscribeTransform;
 import org.apache.beam.sdk.io.kafka.KafkaIO;
+import org.apache.beam.sdk.io.kafka.KafkaSinkMetrics;
 import org.apache.beam.sdk.options.ExperimentalOptions;
 import org.apache.beam.sdk.options.PipelineOptions;
 import org.apache.beam.sdk.options.PipelineOptionsValidator;
@@ -400,6 +401,20 @@ && isServiceEndpoint(dataflowOptions.getDataflowEndpoint())) {
       dataflowOptions.setGcsUploadBufferSizeBytes(GCS_UPLOAD_BUFFER_SIZE_BYTES_DEFAULT);
     }
 
+    boolean hasExperimentEnableKafkaMetrics = false;
+    if (dataflowOptions.getExperiments() != null) {
+      for (String experiment : dataflowOptions.getExperiments()) {
+        if (experiment.startsWith("enable_kafka_metrics")) {
+          hasExperimentEnableKafkaMetrics = true;
+          break;
+        }
+      }
+    }
+
+    if (dataflowOptions.isStreaming() && hasExperimentEnableKafkaMetrics) {
+      KafkaSinkMetrics.setSupportKafkaMetrics(true);
+    }
+
     // Adding the Java version to the SDK name for user's and support convenience.
     String agentJavaVer = "(JRE 8 environment)";
     if (Environments.getJavaVersion() != Environments.JavaVersion.java8) {

diff --git a/...d-platform/src/test/java/org/apache/beam/sdk/io/gcp/bigquery/BigQuerySinkMetricsTest.java b/...d-platform/src/test/java/org/apache/beam/sdk/io/gcp/bigquery/BigQuerySinkMetricsTest.java
@@ -32,10 +32,10 @@
 import java.util.List;
 import java.util.concurrent.ConcurrentHashMap;
 import org.apache.beam.runners.core.metrics.CounterCell;
+import org.apache.beam.runners.core.metrics.HistogramCell;
 import org.apache.beam.runners.core.metrics.MetricsContainerImpl;
 import org.apache.beam.sdk.io.gcp.bigquery.RetryManager.Operation.Context;
 import org.apache.beam.sdk.metrics.Counter;
-import org.apache.beam.sdk.metrics.Histogram;
 import org.apache.beam.sdk.metrics.MetricName;
 import org.apache.beam.sdk.metrics.Metrics;
 import org.apache.beam.sdk.metrics.MetricsEnvironment;
@@ -51,10 +51,14 @@
 @RunWith(JUnit4.class)
 public class BigQuerySinkMetricsTest {
 
-  public static class TestHistogram implements Histogram {
+  public static class TestHistogramCell extends HistogramCell {
     public List<Double> values = Lists.newArrayList();
     private MetricName metricName = MetricName.named("namespace", "name");
 
+    public TestHistogramCell(KV<MetricName, HistogramData.BucketType> kv) {
+      super(kv);
+    }
+
     @Override
     public void update(double value) {
       values.add(value);
@@ -68,10 +72,9 @@ public MetricName getName() {
 
   public static class TestMetricsContainer extends MetricsContainerImpl {
 
-    // public TestHistogram testHistogram = new TestHistogram();
-    public ConcurrentHashMap<KV<MetricName, HistogramData.BucketType>, TestHistogram>
+    public ConcurrentHashMap<KV<MetricName, HistogramData.BucketType>, TestHistogramCell>
         perWorkerHistograms =
-            new ConcurrentHashMap<KV<MetricName, HistogramData.BucketType>, TestHistogram>();
+            new ConcurrentHashMap<KV<MetricName, HistogramData.BucketType>, TestHistogramCell>();
     public ConcurrentHashMap<MetricName, CounterCell> perWorkerCounters =
         new ConcurrentHashMap<MetricName, CounterCell>();
 
@@ -80,11 +83,11 @@ public TestMetricsContainer() {
     }
 
     @Override
-    public Histogram getPerWorkerHistogram(
+    public TestHistogramCell getPerWorkerHistogram(
         MetricName metricName, HistogramData.BucketType bucketType) {
-      perWorkerHistograms.computeIfAbsent(KV.of(metricName, bucketType), kv -> new TestHistogram());
+      perWorkerHistograms.computeIfAbsent(
+          KV.of(metricName, bucketType), kv -> new TestHistogramCell(kv));
       return perWorkerHistograms.get(KV.of(metricName, bucketType));
-      //      return testHistogram;
     }
 
     @Override
@@ -95,7 +98,6 @@ public Counter getPerWorkerCounter(MetricName metricName) {
 
     @Override
     public void reset() {
-      // testHistogram.values.clear();
       perWorkerHistograms.clear();
       perWorkerCounters.clear();
     }
@@ -313,4 +315,4 @@ public void testStreamingInsertsMetrics_enabled() {
         BigQuerySinkMetrics.streamingInsertsMetrics(),
         instanceOf(StreamingInsertsMetrics.StreamingInsertsMetricsImpl.class));
   }
-}
+}
diff --git a/sdks/java/io/kafka/src/main/java/org/apache/beam/sdk/io/kafka/KafkaMetrics.java b/sdks/java/io/kafka/src/main/java/org/apache/beam/sdk/io/kafka/KafkaMetrics.java
@@ -31,8 +31,12 @@
 /** Stores and exports metrics for a batch of Kafka Client RPCs. */
 public interface KafkaMetrics {
 
+  /* Record latency, to be used later to update/create histogram in another thread */
   void updateSuccessfulRpcMetrics(String topic, Duration elapsedTime);
 
+  /* Record and create histogram in current thread */
+  void recordRpcLatencyMetric(String topic, Duration duration);
+
   void updateKafkaMetrics();
 
   /** No-op implementation of {@code KafkaResults}. */
@@ -42,6 +46,9 @@ private NoOpKafkaMetrics() {}
     @Override
     public void updateSuccessfulRpcMetrics(String topic, Duration elapsedTime) {}
 
+    @Override
+    public void recordRpcLatencyMetric(String topic, Duration elapsedTime) {}
+
     @Override
     public void updateKafkaMetrics() {}
 
@@ -78,7 +85,7 @@ public static KafkaMetricsImpl create() {
           new HashMap<String, ConcurrentLinkedQueue<Duration>>(), new AtomicBoolean(true));
     }
 
-    /** Record the rpc status and latency of a successful Kafka poll RPC call. */
+    /** Record the rpc latency of a successful Kafka poll RPC call. */
     @Override
     public void updateSuccessfulRpcMetrics(String topic, Duration elapsedTime) {
       if (isWritable().get()) {
@@ -93,7 +100,7 @@ public void updateSuccessfulRpcMetrics(String topic, Duration elapsedTime) {
       }
     }
 
-    /** Record rpc latency histogram metrics for all recorded topics. */
+    /** Create or update histograms with rpc latency metrics for all recorded topics. */
     private void recordRpcLatencyMetrics() {
       for (Map.Entry<String, ConcurrentLinkedQueue<Duration>> topicLatencies :
           perTopicRpcLatencies().entrySet()) {
@@ -106,14 +113,27 @@ private void recordRpcLatencyMetrics() {
                   KafkaSinkMetrics.RpcMethod.POLL, topicLatencies.getKey());
           latencyHistograms.put(topicLatencies.getKey(), topicHistogram);
         }
-        // update all the latencies
         for (Duration d : topicLatencies.getValue()) {
           Preconditions.checkArgumentNotNull(topicHistogram);
           topicHistogram.update(d.toMillis());
         }
       }
     }
 
+    /** Create or update latency histogram for a singlar topic. */
+    @Override
+    public void recordRpcLatencyMetric(String topic, Duration duration) {
+      Histogram topicHistogram;
+      if (latencyHistograms.containsKey(topic)) {
+        topicHistogram = latencyHistograms.get(topic);
+      } else {
+        topicHistogram =
+            KafkaSinkMetrics.createRPCLatencyHistogram(KafkaSinkMetrics.RpcMethod.POLL, topic);
+        latencyHistograms.put(topic, topicHistogram);
+      }
+      topicHistogram.update(duration.toMillis());
+    }
+
     /**
      * Export all metrics recorded in this instance to the underlying {@code perWorkerMetrics}
      * containers. This function will only report metrics once per instance. Subsequent calls to

diff --git a/sdks/java/io/kafka/src/main/java/org/apache/beam/sdk/io/kafka/KafkaSinkMetrics.java b/sdks/java/io/kafka/src/main/java/org/apache/beam/sdk/io/kafka/KafkaSinkMetrics.java
@@ -34,7 +34,8 @@
 // TODO, refactor out common parts for BQ sink, so it can be reused with other sinks, eg, GCS?
 // @SuppressWarnings("unused")
 public class KafkaSinkMetrics {
-  private static boolean supportKafkaMetrics = false;
+  private static boolean supportKafkaMetrics =
+      true; // where to set to true for UW if experiement is passed
 
   public static final String METRICS_NAMESPACE = "KafkaSink";
 
@@ -50,6 +51,14 @@ enum RpcMethod {
   private static final String TOPIC_LABEL = "topic_name";
   private static final String RPC_METHOD = "rpc_method";
 
+  private static MetricName createMetricName(RpcMethod method, String topic) {
+    LabeledMetricNameUtils.MetricNameBuilder nameBuilder =
+        LabeledMetricNameUtils.MetricNameBuilder.baseNameBuilder(RPC_LATENCY);
+    nameBuilder.addLabel(RPC_METHOD, method.toString());
+    nameBuilder.addLabel(TOPIC_LABEL, topic);
+    return nameBuilder.build(METRICS_NAMESPACE);
+  }
+
   /**
    * Creates an Histogram metric to record RPC latency. Metric will have name.
    *
@@ -60,14 +69,8 @@ enum RpcMethod {
    * @return Histogram with exponential buckets with a sqrt(2) growth factor.
    */
   public static Histogram createRPCLatencyHistogram(RpcMethod method, String topic) {
-    LabeledMetricNameUtils.MetricNameBuilder nameBuilder =
-        LabeledMetricNameUtils.MetricNameBuilder.baseNameBuilder(RPC_LATENCY);
-    nameBuilder.addLabel(RPC_METHOD, method.toString());
-    nameBuilder.addLabel(TOPIC_LABEL, topic);
-
-    MetricName metricName = nameBuilder.build(METRICS_NAMESPACE);
+    MetricName metricName = createMetricName(method, topic);
     HistogramData.BucketType buckets = HistogramData.ExponentialBuckets.of(1, 17);
-
     return new DelegatingHistogram(metricName, buckets, false, true);
   }
 

diff --git a/sdks/java/io/kafka/src/main/java/org/apache/beam/sdk/io/kafka/ReadFromKafkaDoFn.java b/sdks/java/io/kafka/src/main/java/org/apache/beam/sdk/io/kafka/ReadFromKafkaDoFn.java
@@ -228,6 +228,9 @@ private ReadFromKafkaDoFn(
 
   private HashMap<String, Long> perPartitionBacklogMetrics = new HashMap<String, Long>();;
 
+  // Initialize only when used, since its not serializable
+  private @Nullable KafkaMetrics kafkaResults = null;
+
   @VisibleForTesting final long consumerPollingTimeout;
   @VisibleForTesting final DeserializerProvider<K> keyDeserializerProvider;
   @VisibleForTesting final DeserializerProvider<V> valueDeserializerProvider;
@@ -571,7 +574,12 @@ private ConsumerRecords<byte[], byte[]> poll(
     java.time.Duration elapsed = java.time.Duration.ZERO;
     java.time.Duration timeout = java.time.Duration.ofSeconds(this.consumerPollingTimeout);
     while (true) {
+      kafkaResults = KafkaSinkMetrics.kafkaMetrics();
       final ConsumerRecords<byte[], byte[]> rawRecords = consumer.poll(timeout.minus(elapsed));
+      elapsed = sw.elapsed();
+      Preconditions.checkStateNotNull(kafkaResults);
+      kafkaResults.recordRpcLatencyMetric(topicPartition.topic(), elapsed);
+
       if (!rawRecords.isEmpty()) {
         // return as we have found some entries
         return rawRecords;
@@ -580,7 +588,6 @@ private ConsumerRecords<byte[], byte[]> poll(
         // there was no progress on the offset/position, which indicates end of stream
         return rawRecords;
       }
-      elapsed = sw.elapsed();
       if (elapsed.toMillis() >= timeout.toMillis()) {
         // timeout is over
         LOG.warn(

diff --git a/sdks/java/io/kafka/src/test/java/org/apache/beam/sdk/io/kafka/KafkaMetricsTest.java b/sdks/java/io/kafka/src/test/java/org/apache/beam/sdk/io/kafka/KafkaMetricsTest.java
@@ -24,8 +24,8 @@
 import java.time.Duration;
 import java.util.List;
 import java.util.concurrent.ConcurrentHashMap;
+import org.apache.beam.runners.core.metrics.HistogramCell;
 import org.apache.beam.runners.core.metrics.MetricsContainerImpl;
-import org.apache.beam.sdk.metrics.Histogram;
 import org.apache.beam.sdk.metrics.MetricName;
 import org.apache.beam.sdk.metrics.MetricsEnvironment;
 import org.apache.beam.sdk.util.HistogramData;
@@ -39,10 +39,14 @@
 // TODO:Naireen - Refactor to remove duplicate code between the two sinks
 @RunWith(JUnit4.class)
 public class KafkaMetricsTest {
-  public static class TestHistogram implements Histogram {
+  public static class TestHistogramCell extends HistogramCell {
     public List<Double> values = Lists.newArrayList();
     private MetricName metricName = MetricName.named("KafkaSink", "name");
 
+    public TestHistogramCell(KV<MetricName, HistogramData.BucketType> kv) {
+      super(kv);
+    }
+
     @Override
     public void update(double value) {
       values.add(value);
@@ -55,25 +59,21 @@ public MetricName getName() {
   }
 
   public static class TestMetricsContainer extends MetricsContainerImpl {
-    public ConcurrentHashMap<KV<MetricName, HistogramData.BucketType>, TestHistogram>
+    public ConcurrentHashMap<KV<MetricName, HistogramData.BucketType>, TestHistogramCell>
         perWorkerHistograms =
-            new ConcurrentHashMap<KV<MetricName, HistogramData.BucketType>, TestHistogram>();
+            new ConcurrentHashMap<KV<MetricName, HistogramData.BucketType>, TestHistogramCell>();
 
     public TestMetricsContainer() {
       super("TestStep");
     }
 
     @Override
-    public Histogram getPerWorkerHistogram(
+    public TestHistogramCell getPerWorkerHistogram(
         MetricName metricName, HistogramData.BucketType bucketType) {
-      perWorkerHistograms.computeIfAbsent(KV.of(metricName, bucketType), kv -> new TestHistogram());
+      perWorkerHistograms.computeIfAbsent(
+          KV.of(metricName, bucketType), kv -> new TestHistogramCell(kv));
       return perWorkerHistograms.get(KV.of(metricName, bucketType));
     }
-
-    @Override
-    public void reset() {
-      perWorkerHistograms.clear();
-    }
   }
 
   @Test
@@ -126,4 +126,4 @@ public void testKafkaRPCLatencyMetricsAreNotRecorded() throws Exception {
     results.updateKafkaMetrics();
     assertThat(testContainer.perWorkerHistograms.size(), equalTo(0));
   }
-}
+}