KAFKA-18376: High CPU load when AsyncKafkaConsumer uses a small max poll value

clients/src/main/java/org/apache/kafka/clients/consumer/internals/AbstractHeartbeatRequestManager.java

@@ -20,6 +20,7 @@
 import org.apache.kafka.clients.consumer.ConsumerConfig;
 import org.apache.kafka.clients.consumer.internals.NetworkClientDelegate.PollResult;
 import org.apache.kafka.clients.consumer.internals.events.ApplicationEventProcessor;
+import org.apache.kafka.clients.consumer.internals.events.AsyncPollEvent;
 import org.apache.kafka.clients.consumer.internals.events.BackgroundEventHandler;
 import org.apache.kafka.clients.consumer.internals.events.ErrorEvent;
 import org.apache.kafka.clients.consumer.internals.metrics.HeartbeatMetricsManager;
@@ -241,7 +242,7 @@ public PollResult pollOnClose(long currentTimeMs) {
      * are sent, so blocking for longer than the heartbeat interval might mean the application thread is not
      * responsive to changes.
      *
-     * <p>Similarly, we may have to unblock the application thread to send a `PollApplicationEvent` to make sure
+     * <p>Similarly, we may have to unblock the application thread to send a {@link AsyncPollEvent} to make sure
      * our poll timer will not expire while we are polling.
      *
      * <p>In the event that heartbeats are currently being skipped, this still returns the next heartbeat

clients/src/main/java/org/apache/kafka/clients/consumer/internals/ConsumerNetworkThread.java

@@ -20,9 +20,9 @@
 import org.apache.kafka.clients.consumer.internals.events.ApplicationEvent;
 import org.apache.kafka.clients.consumer.internals.events.ApplicationEventProcessor;
 import org.apache.kafka.clients.consumer.internals.events.BackgroundEvent;
-import org.apache.kafka.clients.consumer.internals.events.CompletableApplicationEvent;
 import org.apache.kafka.clients.consumer.internals.events.CompletableEvent;
 import org.apache.kafka.clients.consumer.internals.events.CompletableEventReaper;
+import org.apache.kafka.clients.consumer.internals.events.MetadataErrorNotifiableEvent;
 import org.apache.kafka.clients.consumer.internals.metrics.AsyncConsumerMetrics;
 import org.apache.kafka.common.internals.IdempotentCloser;
 import org.apache.kafka.common.requests.AbstractRequest;
@@ -40,6 +40,7 @@
 import java.util.LinkedList;
 import java.util.List;
 import java.util.Objects;
+import java.util.Optional;
 import java.util.concurrent.BlockingQueue;
 import java.util.function.Supplier;
 
@@ -193,10 +194,13 @@ private void processApplicationEvents() {
             try {
                 if (event instanceof CompletableEvent) {
                     applicationEventReaper.add((CompletableEvent<?>) event);
-                    // Check if there are any metadata errors and fail the CompletableEvent if an error is present.
-                    // This call is meant to handle "immediately completed events" which may not enter the awaiting state,
-                    // so metadata errors need to be checked and handled right away.
-                    maybeFailOnMetadataError(List.of((CompletableEvent<?>) event));
+                }
+                // Check if there are any metadata errors and fail the event if an error is present.
+                // This call is meant to handle "immediately completed events" which may not enter the
+                // awaiting state, so metadata errors need to be checked and handled right away.
+                if (event instanceof MetadataErrorNotifiableEvent) {
+                    if (maybeFailOnMetadataError(List.of(event)))
+                        continue;
                 }
                 applicationEventProcessor.process(event);
             } catch (Throwable t) {
@@ -368,18 +372,26 @@ void cleanup() {
     /**
      * If there is a metadata error, complete all uncompleted events that require subscription metadata.
      */
-    private void maybeFailOnMetadataError(List<CompletableEvent<?>> events) {
-        List<CompletableApplicationEvent<?>> subscriptionMetadataEvent = new ArrayList<>();
+    private boolean maybeFailOnMetadataError(List<?> events) {
+        List<MetadataErrorNotifiableEvent> filteredEvents = new ArrayList<>();
 
-        for (CompletableEvent<?> ce : events) {
-            if (ce instanceof CompletableApplicationEvent && ((CompletableApplicationEvent<?>) ce).requireSubscriptionMetadata())
-                subscriptionMetadataEvent.add((CompletableApplicationEvent<?>) ce);
+        for (Object obj : events) {
+            if (obj instanceof MetadataErrorNotifiableEvent) {
+                filteredEvents.add((MetadataErrorNotifiableEvent) obj);
+            }
         }
 
-        if (subscriptionMetadataEvent.isEmpty())
-            return;
-        networkClientDelegate.getAndClearMetadataError().ifPresent(metadataError ->
-                subscriptionMetadataEvent.forEach(event -> event.future().completeExceptionally(metadataError))
-        );
+        // Don't get-and-clear the metadata error if there are no events that will be notified.
+        if (filteredEvents.isEmpty())
+            return false;
+
+        Optional<Exception> metadataError = networkClientDelegate.getAndClearMetadataError();
+
+        if (metadataError.isPresent()) {
+            filteredEvents.forEach(e -> e.onMetadataError(metadataError.get()));
+            return true;
+        } else {
+            return false;
+        }
     }
 }

clients/src/main/java/org/apache/kafka/clients/consumer/internals/FetchRequestManager.java

@@ -145,6 +145,14 @@ private PollResult pollInternal(FetchRequestPreparer fetchRequestPreparer,
         try {
             Map<Node, FetchSessionHandler.FetchRequestData> fetchRequests = fetchRequestPreparer.prepare();
 
+            if (fetchRequests.isEmpty()) {
+                // If there's nothing to fetch, wake up the FetchBuffer so it doesn't needlessly wait for a wakeup
+                // that won't come until the data in the fetch buffer is consumed.
+                fetchBuffer.wakeup();
+                pendingFetchRequestFuture.complete(null);
+                return PollResult.EMPTY;
+            }
+
             List<UnsentRequest> requests = fetchRequests.entrySet().stream().map(entry -> {
                 final Node fetchTarget = entry.getKey();
                 final FetchSessionHandler.FetchRequestData data = entry.getValue();

clients/src/main/java/org/apache/kafka/clients/consumer/internals/NetworkClientDelegate.java

@@ -471,4 +471,33 @@ protected NetworkClientDelegate create() {
             }
         };
     }
+
+    /**
+     * Creates a {@link Supplier} for deferred creation during invocation by
+     * {@link ConsumerNetworkThread}.
+     */
+    public static Supplier<NetworkClientDelegate> supplier(final Time time,
+                                                           final ConsumerConfig config,
+                                                           final LogContext logContext,
+                                                           final KafkaClient client,
+                                                           final Metadata metadata,
+                                                           final BackgroundEventHandler backgroundEventHandler,
+                                                           final boolean notifyMetadataErrorsViaErrorQueue,
+                                                           final AsyncConsumerMetrics asyncConsumerMetrics) {
+        return new CachedSupplier<>() {
+            @Override
+            protected NetworkClientDelegate create() {
+                return new NetworkClientDelegate(
+                    time,
+                    config,
+                    logContext,
+                    client,
+                    metadata,
+                    backgroundEventHandler,
+                    notifyMetadataErrorsViaErrorQueue,
+                    asyncConsumerMetrics
+                );
+            }
+        };
+    }
 }

clients/src/main/java/org/apache/kafka/clients/consumer/internals/ShareConsumerImpl.java

@@ -38,13 +38,13 @@
 import org.apache.kafka.clients.consumer.internals.events.CompletableEventReaper;
 import org.apache.kafka.clients.consumer.internals.events.ErrorEvent;
 import org.apache.kafka.clients.consumer.internals.events.EventProcessor;
-import org.apache.kafka.clients.consumer.internals.events.PollEvent;
 import org.apache.kafka.clients.consumer.internals.events.ShareAcknowledgeAsyncEvent;
 import org.apache.kafka.clients.consumer.internals.events.ShareAcknowledgeOnCloseEvent;
 import org.apache.kafka.clients.consumer.internals.events.ShareAcknowledgeSyncEvent;
 import org.apache.kafka.clients.consumer.internals.events.ShareAcknowledgementCommitCallbackEvent;
 import org.apache.kafka.clients.consumer.internals.events.ShareAcknowledgementCommitCallbackRegistrationEvent;
 import org.apache.kafka.clients.consumer.internals.events.ShareFetchEvent;
+import org.apache.kafka.clients.consumer.internals.events.SharePollEvent;
 import org.apache.kafka.clients.consumer.internals.events.ShareSubscriptionChangeEvent;
 import org.apache.kafka.clients.consumer.internals.events.ShareUnsubscribeEvent;
 import org.apache.kafka.clients.consumer.internals.events.StopFindCoordinatorOnCloseEvent;
@@ -385,7 +385,7 @@ private void process(final ShareAcknowledgementCommitCallbackEvent event) {
             backgroundEventQueue, time, asyncConsumerMetrics);
 
         final Supplier<NetworkClientDelegate> networkClientDelegateSupplier =
-                () -> new NetworkClientDelegate(time, config, logContext, client, metadata, backgroundEventHandler, true, asyncConsumerMetrics);
+                NetworkClientDelegate.supplier(time, config, logContext, client, metadata, backgroundEventHandler, true, asyncConsumerMetrics);
 
         GroupRebalanceConfig groupRebalanceConfig = new GroupRebalanceConfig(
                 config,
@@ -586,7 +586,7 @@ public synchronized ConsumerRecords<K, V> poll(final Duration timeout) {
 
             do {
                 // Make sure the network thread can tell the application is actively polling
-                applicationEventHandler.add(new PollEvent(timer.currentTimeMs()));
+                applicationEventHandler.add(new SharePollEvent(timer.currentTimeMs()));
 
                 processBackgroundEvents();
 

clients/src/main/java/org/apache/kafka/clients/consumer/internals/StreamsGroupHeartbeatRequestManager.java

@@ -18,6 +18,7 @@
 
 import org.apache.kafka.clients.CommonClientConfigs;
 import org.apache.kafka.clients.consumer.ConsumerConfig;
+import org.apache.kafka.clients.consumer.internals.events.AsyncPollEvent;
 import org.apache.kafka.clients.consumer.internals.events.BackgroundEventHandler;
 import org.apache.kafka.clients.consumer.internals.events.ErrorEvent;
 import org.apache.kafka.clients.consumer.internals.metrics.HeartbeatMetricsManager;
@@ -426,7 +427,7 @@ public StreamsMembershipManager membershipManager() {
      * are sent, so blocking for longer than the heartbeat interval might mean the application thread is not
      * responsive to changes.
      *
-     * <p>Similarly, we may have to unblock the application thread to send a `PollApplicationEvent` to make sure
+     * <p>Similarly, we may have to unblock the application thread to send a {@link AsyncPollEvent} to make sure
      * our poll timer will not expire while we are polling.
      *
      * <p>In the event that heartbeats are currently being skipped, this still returns the next heartbeat

clients/src/main/java/org/apache/kafka/clients/consumer/internals/events/AbstractTopicMetadataEvent.java

@@ -21,14 +21,14 @@
 import java.util.List;
 import java.util.Map;
 
-public abstract class AbstractTopicMetadataEvent extends CompletableApplicationEvent<Map<String, List<PartitionInfo>>> {
+public abstract class AbstractTopicMetadataEvent extends CompletableApplicationEvent<Map<String, List<PartitionInfo>>> implements MetadataErrorNotifiableEvent {
 
     protected AbstractTopicMetadataEvent(final Type type, final long deadlineMs) {
         super(type, deadlineMs);
     }
 
     @Override
-    public boolean requireSubscriptionMetadata() {
-        return true;
+    public void onMetadataError(Exception metadataError) {
+        future().completeExceptionally(metadataError);
     }
 }

clients/src/main/java/org/apache/kafka/clients/consumer/internals/events/ApplicationEvent.java

@@ -28,14 +28,14 @@
 public abstract class ApplicationEvent {
 
     public enum Type {
-        COMMIT_ASYNC, COMMIT_SYNC, POLL, FETCH_COMMITTED_OFFSETS, NEW_TOPICS_METADATA_UPDATE, ASSIGNMENT_CHANGE,
+        COMMIT_ASYNC, COMMIT_SYNC, ASYNC_POLL, FETCH_COMMITTED_OFFSETS, NEW_TOPICS_METADATA_UPDATE, ASSIGNMENT_CHANGE,
         LIST_OFFSETS, CHECK_AND_UPDATE_POSITIONS, RESET_OFFSET, TOPIC_METADATA, ALL_TOPICS_METADATA,
         TOPIC_SUBSCRIPTION_CHANGE, TOPIC_PATTERN_SUBSCRIPTION_CHANGE, TOPIC_RE2J_PATTERN_SUBSCRIPTION_CHANGE,
         UPDATE_SUBSCRIPTION_METADATA, UNSUBSCRIBE,
         CONSUMER_REBALANCE_LISTENER_CALLBACK_COMPLETED,
         COMMIT_ON_CLOSE, CREATE_FETCH_REQUESTS, LEAVE_GROUP_ON_CLOSE, STOP_FIND_COORDINATOR_ON_CLOSE,
         PAUSE_PARTITIONS, RESUME_PARTITIONS, CURRENT_LAG,
-        SHARE_FETCH, SHARE_ACKNOWLEDGE_ASYNC, SHARE_ACKNOWLEDGE_SYNC,
+        SHARE_POLL, SHARE_FETCH, SHARE_ACKNOWLEDGE_ASYNC, SHARE_ACKNOWLEDGE_SYNC,
         SHARE_SUBSCRIPTION_CHANGE, SHARE_UNSUBSCRIBE,
         SHARE_ACKNOWLEDGE_ON_CLOSE,
         SHARE_ACKNOWLEDGEMENT_COMMIT_CALLBACK_REGISTRATION,