diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/wal/AsyncFSWAL.java b/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/wal/AsyncFSWAL.java
index 7c8b3794597..5699b3d2f99 100644
--- a/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/wal/AsyncFSWAL.java
+++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/regionserver/wal/AsyncFSWAL.java
@@ -437,7 +437,11 @@ public class AsyncFSWAL extends AbstractFSWAL<AsyncWriter> {
       newHighestProcessedAppendTxid = entry.getTxid();
       iter.remove();
       if (appended) {
-        unackedAppends.addLast(entry);
+        // This is possible, when we fail to sync, we will add the unackedAppends back to
+        // toWriteAppends, so here we may get an entry which is already in the unackedAppends.
+        if (unackedAppends.isEmpty() || unackedAppends.peekLast().getTxid() < entry.getTxid()) {
+          unackedAppends.addLast(entry);
+        }
         if (writer.getLength() - fileLengthAtLastSync >= batchSize) {
           break;
         }