diff --git a/src/main/java/io/vertx/kafka/client/consumer/impl/KafkaReadStreamImpl.java b/src/main/java/io/vertx/kafka/client/consumer/impl/KafkaReadStreamImpl.java
index 70f0ad3e..6c04805b 100644
--- a/src/main/java/io/vertx/kafka/client/consumer/impl/KafkaReadStreamImpl.java
+++ b/src/main/java/io/vertx/kafka/client/consumer/impl/KafkaReadStreamImpl.java
@@ -62,6 +62,7 @@ public class KafkaReadStreamImpl<K, V> implements KafkaReadStream<K, V> {
 
   private final AtomicBoolean consuming = new AtomicBoolean(false);
   private final AtomicLong demand = new AtomicLong(Long.MAX_VALUE);
+  private final AtomicBoolean polling = new AtomicBoolean(false);
   private Handler<ConsumerRecord<K, V>> recordHandler;
   private Handler<Throwable> exceptionHandler;
   private Iterator<ConsumerRecord<K, V>> current; // Accessed on event loop
@@ -137,24 +138,37 @@ private <T> void submitTaskWhenStarted(java.util.function.BiConsumer<Consumer<K,
   }
 
   private void pollRecords(Handler<ConsumerRecords<K, V>> handler) {
-    this.worker.submit(() -> {
-      if (!this.closed.get()) {
-        try {
-          ConsumerRecords<K, V> records = this.consumer.poll(pollTimeout);
-          if (records != null && records.count() > 0) {
-            this.context.runOnContext(v -> handler.handle(records));
-          } else {
-            // Don't call pollRecords directly, but use schedule() to actually pause when the readStream is paused
-            schedule(0);
-          }
-        } catch (WakeupException ignore) {
-        } catch (Exception e) {
-          if (exceptionHandler != null) {
-            exceptionHandler.handle(e);
-          }
-        }
+      if(this.polling.compareAndSet(false, true)){
+          this.worker.submit(() -> {
+             boolean submitted = false;
+             try {
+                if (!this.closed.get()) {
+                  try {
+                    ConsumerRecords<K, V> records = this.consumer.poll(pollTimeout);
+                    if (records != null && records.count() > 0) {
+                      submitted = true; // sets false only when the iterator is overwritten
+                      this.context.runOnContext(v -> {
+                          this.polling.set(false);
+                          handler.handle(records);
+                      });
+                    }
+                  } catch (WakeupException ignore) {
+                  } catch (Exception e) {
+                    if (exceptionHandler != null) {
+                      exceptionHandler.handle(e);
+                    }
+                  }
+                }
+             } finally {
+                 if(!submitted){
+                     this.context.runOnContext(v -> {
+                         this.polling.set(false);
+                         schedule(0);
+                     });
+                 }
+             }
+          });
       }
-    });
   }
 
   private void schedule(long delay) {
diff --git a/src/test/java/io/vertx/kafka/client/tests/KafkaReadStreamMockTest.java b/src/test/java/io/vertx/kafka/client/tests/KafkaReadStreamMockTest.java
new file mode 100644
index 00000000..6262ea9e
--- /dev/null
+++ b/src/test/java/io/vertx/kafka/client/tests/KafkaReadStreamMockTest.java
@@ -0,0 +1,120 @@
+package io.vertx.kafka.client.tests;
+
+import java.util.HashMap;
+import java.util.LinkedHashSet;
+import java.util.LinkedList;
+import java.util.Map;
+import java.util.Set;
+import java.util.concurrent.atomic.AtomicLong;
+
+import org.apache.kafka.clients.consumer.ConsumerRecord;
+import org.apache.kafka.clients.consumer.MockConsumer;
+import org.apache.kafka.clients.consumer.OffsetResetStrategy;
+import org.junit.Test;
+import org.junit.runner.RunWith;
+
+import io.vertx.core.Vertx;
+import io.vertx.ext.unit.Async;
+import io.vertx.ext.unit.TestContext;
+import io.vertx.ext.unit.junit.VertxUnitRunner;
+import io.vertx.kafka.client.common.TopicPartition;
+import io.vertx.kafka.client.consumer.KafkaConsumer;
+import io.vertx.kafka.client.consumer.KafkaReadStream;
+import io.vertx.kafka.client.consumer.impl.KafkaConsumerImpl;
+import io.vertx.kafka.client.producer.KafkaProducerRecord;
+import io.vertx.kafka.client.producer.impl.KafkaProducerRecordImpl;
+
+@RunWith(VertxUnitRunner.class)
+public class KafkaReadStreamMockTest extends KafkaTestBase {
+
+    private LinkedList<ConsumerRecord<String,String>> recordsMock = new LinkedList<>();
+    
+    private int SEND_BATCH = 5;
+    private int TOTAL_MESSAGES = 400;
+    private final String TOPIC = "topic";
+    private Long timer = null;
+    
+    private void initRecords(){
+        int numMessages = TOTAL_MESSAGES;
+        for (int i = 0;i < numMessages;i++) {
+            String key = "key-" + i;
+            String value = "value-" + i;
+            recordsMock.add(new ConsumerRecord<String,String>(TOPIC, 0, i, key, value));
+        } 
+    }
+    
+    private MockConsumer<String, String> createMockConsumer(){
+        MockConsumer<String, String> consumer = new MockConsumer<>(OffsetResetStrategy.EARLIEST);
+
+        Map<org.apache.kafka.common.TopicPartition, Long> beginningOffsets = new HashMap<>();
+        beginningOffsets.put(new org.apache.kafka.common.TopicPartition(TOPIC, 0), 0L);
+        consumer.updateBeginningOffsets(beginningOffsets);
+        return consumer;
+    }
+    
+    private void sendNextBatch(MockConsumer<String, String> consumer){
+        for(int i=0;i<SEND_BATCH && recordsMock.size()>0;i++)
+        consumer.addRecord(recordsMock.pop());
+        
+    }
+    
+    @Test
+    public void shouldNotLoseMessages(TestContext ctx){
+        Vertx vertx = Vertx.vertx();
+        
+        Async done = ctx.async();
+        
+        initRecords();
+
+        MockConsumer<String, String> consumer = createMockConsumer();
+        KafkaReadStream<String, String> readStream =  KafkaReadStream.create(vertx, consumer);
+        KafkaConsumer<String, String> consumerVertx = new KafkaConsumerImpl<>(readStream);
+        
+        
+        AtomicLong partitionOffset = new AtomicLong(-1);
+        
+        
+        consumerVertx.handler((r)->{
+            long offset = r.offset();
+            
+            partitionOffset.addAndGet(1);
+            ctx.assertEquals(partitionOffset.get(), offset);
+            
+            System.out.println("offset " + offset);
+            
+            if(offset == TOTAL_MESSAGES-1){
+                consumerVertx.close();
+                done.complete();
+            } else {
+                
+                if(timer!=null) vertx.cancelTimer(timer);
+                timer = vertx.setTimer(5, (t)->{
+                    consumerVertx.pause();
+                    vertx.getOrCreateContext().runOnContext((t1)->{
+                        consumerVertx.commit();
+                        consumerVertx.resume();
+                        sendNextBatch(consumer);
+                        // sends two batches of messages
+                        vertx.getOrCreateContext().runOnContext((t2)->{
+                            sendNextBatch(consumer);
+                        });
+                    });
+                });
+            
+            }
+
+        });
+        
+        consumerVertx.exceptionHandler(t->ctx.fail(t));
+
+        
+        Set<TopicPartition> partitions = new LinkedHashSet<>();
+        partitions.add(new TopicPartition(TOPIC, 0));
+        
+        consumerVertx.assign(partitions, (h)->{
+            sendNextBatch(consumer);
+        });
+        
+    }
+    
+}