diff --git a/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt b/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
index df01c05d8d1..1c0a5a3f09e 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
+++ b/hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt
@@ -341,6 +341,9 @@ Release 2.0.3-alpha - Unreleased
     HDFS-4308. addBlock() should persist file blocks once.
     (Plamen Jeliazkov via shv)
 
+    HDFS-4347. Avoid infinite waiting checkpoint to complete in TestBackupNode.
+    (Plamen Jeliazkov via shv)
+
   BREAKDOWN OF HDFS-3077 SUBTASKS
 
     HDFS-3077. Quorum-based protocol for reading and writing edit logs.
diff --git a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestBackupNode.java b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestBackupNode.java
index 17192eb8044..b19645d0392 100644
--- a/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestBackupNode.java
+++ b/hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestBackupNode.java
@@ -325,8 +325,8 @@ public class TestBackupNode {
       //
       // Take a checkpoint
       //
-      backup = startBackupNode(conf, op, 1);
       long txid = cluster.getNameNodeRpc().getTransactionID();
+      backup = startBackupNode(conf, op, 1);
       waitCheckpointDone(cluster, txid);
 
       for (int i = 0; i < 10; i++) {