HDFS-9530. ReservedSpace is not cleared for abandoned Blocks (Contributed by Brahma Reddy Battula)

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/DataNodeFaultInjector.java

@@ -53,4 +53,6 @@ public boolean dropHeartbeatPacket() {
   public void stopSendingPacketDownstream() throws IOException {}
 
   public void noRegistration() throws IOException { }
+
+  public void failMirrorConnection() throws IOException { }
 }

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/DataXceiver.java

@@ -739,6 +739,9 @@ public void writeBlock(final ExtendedBlock block,
         mirrorTarget = NetUtils.createSocketAddr(mirrorNode);
         mirrorSock = datanode.newSocket();
         try {
+
+          DataNodeFaultInjector.get().failMirrorConnection();
+
           int timeoutValue = dnConf.socketTimeout +
               (HdfsConstants.READ_TIMEOUT_EXTENSION * targets.length);
           int writeTimeout = dnConf.socketWriteTimeout +

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/FsDatasetImpl.java

@@ -79,6 +79,7 @@
 import org.apache.hadoop.hdfs.server.datanode.ReplicaBeingWritten;
 import org.apache.hadoop.hdfs.server.datanode.ReplicaHandler;
 import org.apache.hadoop.hdfs.server.datanode.ReplicaInPipeline;
+import org.apache.hadoop.hdfs.server.datanode.ReplicaInPipelineInterface;
 import org.apache.hadoop.hdfs.server.datanode.ReplicaInfo;
 import org.apache.hadoop.hdfs.server.datanode.ReplicaNotFoundException;
 import org.apache.hadoop.hdfs.server.datanode.ReplicaUnderRecovery;
@@ -1955,6 +1956,9 @@ public void invalidate(String bpid, Block invalidBlks[]) throws IOException {
           LOG.debug("Block file " + removing.getBlockFile().getName()
               + " is to be deleted");
         }
+        if (removing instanceof ReplicaInPipelineInterface) {
+          ((ReplicaInPipelineInterface) removing).releaseAllBytesReserved();
+        }
       }
 
       if (v.isTransientStorage()) {

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/datanode/fsdataset/impl/TestSpaceReservation.java

@@ -36,6 +36,7 @@
 import org.apache.hadoop.hdfs.*;
 import org.apache.hadoop.hdfs.protocol.Block;
 import org.apache.hadoop.hdfs.server.datanode.DataNode;
+import org.apache.hadoop.hdfs.server.datanode.DataNodeFaultInjector;
 import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsDatasetSpi;
 import org.apache.hadoop.hdfs.server.datanode.fsdataset.FsVolumeReference;
 import org.apache.hadoop.ipc.RemoteException;
@@ -77,6 +78,7 @@ public class TestSpaceReservation {
   private DFSClient client = null;
   FsVolumeReference singletonVolumeRef = null;
   FsVolumeImpl singletonVolume = null;
+  private DataNodeFaultInjector old = null;
 
   private static Random rand = new Random();
 
@@ -146,6 +148,9 @@ public void shutdownCluster() throws IOException {
       cluster.shutdown();
       cluster = null;
     }
+    if (old != null) {
+      DataNodeFaultInjector.set(old);
+    }
   }
 
   private void createFileAndTestSpaceReservation(
@@ -613,6 +618,49 @@ public void testReservedSpaceForAppend() throws Exception {
     checkReservedSpace(expectedFile2Reserved);
   }
 
+  @Test(timeout = 30000)
+  public void testReservedSpaceForPipelineRecovery() throws Exception {
+    final short replication = 3;
+    startCluster(BLOCK_SIZE, replication, -1);
+
+    final String methodName = GenericTestUtils.getMethodName();
+    final Path file = new Path("/" + methodName + ".01.dat");
+
+    old = DataNodeFaultInjector.get();
+    // Fault injector to fail connection to mirror first time.
+    DataNodeFaultInjector.set(new DataNodeFaultInjector() {
+      private int tries = 0;
+
+      @Override
+      public void failMirrorConnection() throws IOException {
+        if (tries++ == 0) {
+          throw new IOException("Failing Mirror for space reservation");
+        }
+      }
+    });
+    // Write 1 byte to the file and kill the writer.
+    FSDataOutputStream os = fs.create(file, replication);
+    os.write(new byte[1]);
+    os.close();
+    // Ensure all space reserved for the replica was released on each
+    // DataNode.
+    cluster.triggerBlockReports();
+    for (final DataNode dn : cluster.getDataNodes()) {
+      try (FsDatasetSpi.FsVolumeReferences volumes =
+          dn.getFSDataset().getFsVolumeReferences()) {
+        final FsVolumeImpl volume = (FsVolumeImpl) volumes.get(0);
+        GenericTestUtils.waitFor(new Supplier<Boolean>() {
+          @Override
+          public Boolean get() {
+            LOG.info("dn " + dn.getDisplayName() + " space : "
+                + volume.getReservedForReplicas());
+            return (volume.getReservedForReplicas() == 0);
+          }
+        }, 100, Integer.MAX_VALUE); // Wait until the test times out.
+      }
+    }
+  }
+
   private void checkReservedSpace(final long expectedReserved) throws TimeoutException,
       InterruptedException, IOException {
     for (final DataNode dn : cluster.getDataNodes()) {