diff --git a/CHANGELOG.md b/CHANGELOG.md
index ae045fc560a..1729d91d21f 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -7,6 +7,7 @@
 * Fix #4793: (java-generator) Fix broken POJO generation when two schema properties collide into a single field name
 * Fix #4963: Openshift Client return 403 when use websocket
 * Fix #4985: triggering the immediate cleanup of the okhttp idle task
+* Fix #4988: Ensuring that previous requests are closed before retry
 * fix #5002: Jetty response completion accounts for header processing
 * Fix #5009: addressing issue with serialization of wrapped polymophic types
 
diff --git a/kubernetes-client-api/src/main/java/io/fabric8/kubernetes/client/http/StandardHttpClient.java b/kubernetes-client-api/src/main/java/io/fabric8/kubernetes/client/http/StandardHttpClient.java
index 917e91e92f0..da24bcb55e5 100644
--- a/kubernetes-client-api/src/main/java/io/fabric8/kubernetes/client/http/StandardHttpClient.java
+++ b/kubernetes-client-api/src/main/java/io/fabric8/kubernetes/client/http/StandardHttpClient.java
@@ -145,6 +145,7 @@ protected <V> void retryWithExponentialBackoff(CompletableFuture<V> result,
                 LOG.debug("HTTP operation on url: {} should be retried as the response code was {}, retrying after {} millis",
                     uri, code, retryInterval);
                 retry = true;
+                cancel.accept(response);
               }
             } else if (throwable instanceof IOException) {
               LOG.debug(String.format("HTTP operation on url: %s should be retried after %d millis because of IOException",
diff --git a/kubernetes-client-api/src/test/java/io/fabric8/kubernetes/client/http/StandardHttpClientTest.java b/kubernetes-client-api/src/test/java/io/fabric8/kubernetes/client/http/StandardHttpClientTest.java
index 5e308c28bac..81ddbaccc5f 100644
--- a/kubernetes-client-api/src/test/java/io/fabric8/kubernetes/client/http/StandardHttpClientTest.java
+++ b/kubernetes-client-api/src/test/java/io/fabric8/kubernetes/client/http/StandardHttpClientTest.java
@@ -143,7 +143,8 @@ void testHttpRetryWithLessFailuresThanRetries() throws Exception {
         .withRequestRetryBackoffInterval(50).build())
         .build();
 
-    final HttpResponse<AsyncBody> error = new TestHttpResponse<AsyncBody>().withCode(500);
+    final HttpResponse<AsyncBody> error = new TestHttpResponse<AsyncBody>().withBody(Mockito.mock(AsyncBody.class))
+        .withCode(500);
     client.getRespFutures().add(CompletableFuture.completedFuture(error));
     client.getRespFutures().add(CompletableFuture.completedFuture(error));
     client.getRespFutures().add(CompletableFuture.completedFuture(error));
@@ -155,7 +156,7 @@ void testHttpRetryWithLessFailuresThanRetries() throws Exception {
         });
 
     // should ultimately succeed with the final 200
-    assertEquals(200, consumeFuture.get().code());
+    assertEquals(200, consumeFuture.get(2, TimeUnit.MINUTES).code());
 
     // only 4 requests issued
     assertEquals(4, client.getRespFutures().size());
@@ -179,9 +180,35 @@ void testWebSocketWithLessFailuresThanRetries() throws Exception {
     client.getWsFutures().add(client.getWsFutures().get(0));
     client.getWsFutures().add(CompletableFuture.completedFuture((new WebSocketResponse(ws, null))));
 
-    future.get();
+    future.get(2, TimeUnit.MINUTES);
 
     assertEquals(3, client.getWsFutures().size());
   }
 
+  @Test
+  void testClosePreviousBeforeRetry() throws Exception {
+    client = client.newBuilder().tag(new RequestConfigBuilder()
+        .withRequestRetryBackoffLimit(1)
+        .withRequestRetryBackoffInterval(50).build())
+        .build();
+
+    final HttpResponse<AsyncBody> error = new TestHttpResponse<AsyncBody>().withBody(Mockito.mock(AsyncBody.class))
+        .withCode(503);
+    client.getRespFutures().add(CompletableFuture.completedFuture(error));
+    client.getRespFutures().add(CompletableFuture.completedFuture(new TestHttpResponse<AsyncBody>().withCode(200)));
+
+    CompletableFuture<HttpResponse<AsyncBody>> consumeFuture = client.consumeBytes(
+        client.newHttpRequestBuilder().uri("http://localhost").build(),
+        (value, asyncBody) -> {
+        });
+
+    Mockito.verify(error.body()).cancel();
+
+    // should ultimately succeed with the final 200
+    assertEquals(200, consumeFuture.get(2, TimeUnit.MINUTES).code());
+
+    // only 2 requests issued
+    assertEquals(2, client.getRespFutures().size());
+  }
+
 }