HDFS-8308. Erasure Coding: NameNode may get blocked in waitForLoadingFSImage() when loading editlog. Contributed by Jing Zhao.

hadoop-hdfs-project/hadoop-hdfs/CHANGES-HDFS-EC-7285.txt

@@ -152,3 +152,6 @@
 
     HDFS-8183. Erasure Coding: Improve DFSStripedOutputStream closing of 
     datastreamer threads. (Rakesh R via Zhe Zhang)
+
+    HDFS-8308. Erasure Coding: NameNode may get blocked in waitForLoadingFSImage()
+    when loading editlog. (jing9)

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/ErasureCodingZoneManager.java

@@ -79,7 +79,8 @@ ECZoneInfo getECZoneInfo(INodesInPath iip) throws IOException {
       for (XAttr xAttr : xAttrs) {
         if (XATTR_ERASURECODING_ZONE.equals(XAttrHelper.getPrefixName(xAttr))) {
           String schemaName = new String(xAttr.getValue());
-          ECSchema schema = dir.getFSNamesystem().getECSchema(schemaName);
+          ECSchema schema = dir.getFSNamesystem().getSchemaManager()
+              .getSchema(schemaName);
           return new ECZoneInfo(inode.getFullPathName(), schema);
         }
       }

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/namenode/FSNamesystem.java

@@ -7529,9 +7529,9 @@ BatchedListEntries<EncryptionZone> listEncryptionZones(long prevId)
 
   /**
    * Create an erasure coding zone on directory src.
-   * @param schema  ECSchema for the erasure coding zone
+   * @param srcArg  the path of a directory which will be the root of the
-   * @param src     the path of a directory which will be the root of the
    *                erasure coding zone. The directory must be empty.
+   * @param schema  ECSchema for the erasure coding zone
    *
    * @throws AccessControlException  if the caller is not the superuser.
    * @throws UnresolvedLinkException if the path can't be resolved.

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/DFSTestUtil.java

@@ -1948,4 +1948,16 @@ public static Block addStripedBlockToFile(List<DataNode> dataNodes,
     lastBlock.setNumBytes(numStripes * BLOCK_STRIPED_CELL_SIZE * NUM_DATA_BLOCKS);
     return lastBlock;
   }
+
+  /**
+   * Because currently DFSStripedOutputStream does not support hflush/hsync,
+   * tests can use this method to flush all the buffered data to DataNodes.
+   */
+  public static void writeAndFlushStripedOutputStream(
+      DFSStripedOutputStream out, int chunkSize) throws IOException {
+    // FSOutputSummer.BUFFER_NUM_CHUNKS == 9
+    byte[] toWrite = new byte[chunkSize * 9 + 1];
+    out.write(toWrite);
+    out.flushInternal();
+  }
 }

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/TestErasureCodingZones.java

@@ -35,8 +35,6 @@
 import static org.junit.Assert.*;
 
 public class TestErasureCodingZones {
-  private final int NUM_OF_DATANODES = 3;
-  private Configuration conf;
   private MiniDFSCluster cluster;
   private DistributedFileSystem fs;
   private static final int BLOCK_SIZE = 1024;
@@ -44,10 +42,10 @@ public class TestErasureCodingZones {
 
   @Before
   public void setupCluster() throws IOException {
-    conf = new HdfsConfiguration();
+    Configuration conf = new HdfsConfiguration();
     conf.setInt(DFSConfigKeys.DFS_BLOCK_SIZE_KEY, BLOCK_SIZE);
     cluster = new MiniDFSCluster.Builder(conf).
-        numDataNodes(NUM_OF_DATANODES).build();
+        numDataNodes(1).build();
     cluster.waitActive();
     fs = cluster.getFileSystem();
     namesystem = cluster.getNamesystem();

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/namenode/TestAddStripedBlocks.java

@@ -19,6 +19,7 @@
 
 import org.apache.hadoop.fs.FSDataOutputStream;
 import org.apache.hadoop.fs.Path;
+import org.apache.hadoop.hdfs.DFSStripedOutputStream;
 import org.apache.hadoop.hdfs.DFSTestUtil;
 import org.apache.hadoop.hdfs.DistributedFileSystem;
 import org.apache.hadoop.hdfs.HdfsConfiguration;
@@ -55,6 +56,7 @@
 import java.util.List;
 import java.util.UUID;
 
+import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_BYTES_PER_CHECKSUM_DEFAULT;
 import static org.apache.hadoop.hdfs.protocol.HdfsConstants.BLOCK_STRIPED_CELL_SIZE;
 import static org.apache.hadoop.hdfs.protocol.HdfsConstants.NUM_DATA_BLOCKS;
 import static org.junit.Assert.assertEquals;
@@ -103,36 +105,31 @@ public void testAllocateBlockId() throws Exception {
     Assert.assertEquals(firstId + HdfsConstants.MAX_BLOCKS_IN_GROUP, secondId);
   }
 
-  @Test
+  @Test (timeout=60000)
   public void testAddStripedBlock() throws Exception {
     final Path file = new Path("/file1");
     // create an empty file
     FSDataOutputStream out = null;
     try {
       out = dfs.create(file, (short) 1);
+      DFSTestUtil.writeAndFlushStripedOutputStream(
+          (DFSStripedOutputStream) out.getWrappedStream(),
+          DFS_BYTES_PER_CHECKSUM_DEFAULT);
 
       FSDirectory fsdir = cluster.getNamesystem().getFSDirectory();
       INodeFile fileNode = fsdir.getINode4Write(file.toString()).asFile();
-      LocatedBlock newBlock = cluster.getNamesystem().getAdditionalBlock(
-          file.toString(), fileNode.getId(), dfs.getClient().getClientName(),
-          null, null, null);
-      assertEquals(GROUP_SIZE, newBlock.getLocations().length);
-      assertEquals(GROUP_SIZE, newBlock.getStorageIDs().length);
 
       BlockInfo[] blocks = fileNode.getBlocks();
       assertEquals(1, blocks.length);
       Assert.assertTrue(blocks[0].isStriped());
 
       checkStripedBlockUC((BlockInfoStriped) fileNode.getLastBlock(), true);
-    } finally {
-      IOUtils.cleanup(null, out);
-    }
 
       // restart NameNode to check editlog
       cluster.restartNameNode(true);
-    FSDirectory fsdir = cluster.getNamesystem().getFSDirectory();
+      fsdir = cluster.getNamesystem().getFSDirectory();
-    INodeFile fileNode = fsdir.getINode4Write(file.toString()).asFile();
+      fileNode = fsdir.getINode4Write(file.toString()).asFile();
-    BlockInfo[] blocks = fileNode.getBlocks();
+      blocks = fileNode.getBlocks();
       assertEquals(1, blocks.length);
       Assert.assertTrue(blocks[0].isStriped());
       checkStripedBlockUC((BlockInfoStriped) fileNode.getLastBlock(), false);
@@ -149,6 +146,9 @@ public void testAddStripedBlock() throws Exception {
       assertEquals(1, blocks.length);
       Assert.assertTrue(blocks[0].isStriped());
       checkStripedBlockUC((BlockInfoStriped) fileNode.getLastBlock(), false);
+    } finally {
+      IOUtils.cleanup(null, out);
+    }
   }
 
   private void checkStripedBlockUC(BlockInfoStriped block,
@@ -190,11 +190,12 @@ public void testGetLocatedStripedBlocks() throws Exception {
     FSDataOutputStream out = null;
     try {
       out = dfs.create(file, (short) 1);
+      DFSTestUtil.writeAndFlushStripedOutputStream(
+          (DFSStripedOutputStream) out.getWrappedStream(),
+          DFS_BYTES_PER_CHECKSUM_DEFAULT);
 
       FSDirectory fsdir = cluster.getNamesystem().getFSDirectory();
       INodeFile fileNode = fsdir.getINode4Write(file.toString()).asFile();
-      cluster.getNamesystem().getAdditionalBlock(file.toString(),
-          fileNode.getId(), dfs.getClient().getClientName(), null, null, null);
       BlockInfoStripedUnderConstruction lastBlk =
           (BlockInfoStripedUnderConstruction) fileNode.getLastBlock();
       DatanodeInfo[] expectedDNs = DatanodeStorageInfo