diff --git a/core/src/main/scala/org/apache/spark/deploy/master/Master.scala b/core/src/main/scala/org/apache/spark/deploy/master/Master.scala
index f8aac3008cef..fded8475a091 100644
--- a/core/src/main/scala/org/apache/spark/deploy/master/Master.scala
+++ b/core/src/main/scala/org/apache/spark/deploy/master/Master.scala
@@ -265,19 +265,16 @@ private[deploy] class Master(
 
             val normalExit = exitStatus == Some(0)
             // Only retry certain number of times so we don't go into an infinite loop.
-            if (!normalExit) {
-              if (appInfo.incrementRetryCount() < ApplicationState.MAX_NUM_RETRY) {
-                schedule()
-              } else {
-                val execs = appInfo.executors.values
-                if (!execs.exists(_.state == ExecutorState.RUNNING)) {
-                  logError(s"Application ${appInfo.desc.name} with ID ${appInfo.id} failed " +
-                    s"${appInfo.retryCount} times; removing it")
-                  removeApplication(appInfo, ApplicationState.FAILED)
-                }
+            if (!normalExit && appInfo.incrementRetryCount() >= ApplicationState.MAX_NUM_RETRY) {
+              val execs = appInfo.executors.values
+              if (!execs.exists(_.state == ExecutorState.RUNNING)) {
+                logError(s"Application ${appInfo.desc.name} with ID ${appInfo.id} failed " +
+                  s"${appInfo.retryCount} times; removing it")
+                removeApplication(appInfo, ApplicationState.FAILED)
               }
             }
           }
+          schedule()
         case None =>
           logWarning(s"Got status update for unknown executor $appId/$execId")
       }
diff --git a/core/src/test/scala/org/apache/spark/DistributedSuite.scala b/core/src/test/scala/org/apache/spark/DistributedSuite.scala
index 0515e6e3a631..6beae842b04d 100644
--- a/core/src/test/scala/org/apache/spark/DistributedSuite.scala
+++ b/core/src/test/scala/org/apache/spark/DistributedSuite.scala
@@ -134,6 +134,21 @@ class DistributedSuite extends SparkFunSuite with Matchers with LocalSparkContex
     }
   }
 
+  test("repeatedly failing task that crashes JVM with a zero exit code (SPARK-16925)") {
+    // Ensures that if a task which causes the JVM to exit with a zero exit code will cause the
+    // Spark job to eventually fail.
+    sc = new SparkContext(clusterUrl, "test")
+    failAfter(Span(100000, Millis)) {
+      val thrown = intercept[SparkException] {
+        sc.parallelize(1 to 1, 1).foreachPartition { _ => System.exit(0) }
+      }
+      assert(thrown.getClass === classOf[SparkException])
+      assert(thrown.getMessage.contains("failed 4 times"))
+    }
+    // Check that the cluster is still usable:
+    sc.parallelize(1 to 10).count()
+  }
+
   test("caching") {
     sc = new SparkContext(clusterUrl, "test")
     val data = sc.parallelize(1 to 1000, 10).cache()