Package org.apache.hadoop.hdfs.server.blockmanagement

Source Code of org.apache.hadoop.hdfs.server.blockmanagement.TestBlockManager

/**
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements.  See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership.  The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License.  You may obtain a copy of the License at
*
*     http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hadoop.hdfs.server.blockmanagement;

import static org.junit.Assert.assertEquals;
import static org.junit.Assert.assertFalse;
import static org.junit.Assert.assertNotNull;
import static org.junit.Assert.assertNull;
import static org.junit.Assert.assertTrue;
import static org.mockito.Mockito.doReturn;
import static org.mockito.Mockito.reset;
import static org.mockito.Mockito.spy;
import static org.mockito.Mockito.verify;

import java.io.IOException;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.LinkedList;
import java.util.List;
import java.util.Map.Entry;

import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.hdfs.DFSConfigKeys;
import org.apache.hadoop.hdfs.DFSTestUtil;
import org.apache.hadoop.hdfs.HdfsConfiguration;
import org.apache.hadoop.hdfs.protocol.Block;
import org.apache.hadoop.hdfs.protocol.BlockListAsLongs;
import org.apache.hadoop.hdfs.protocol.HdfsConstants;
import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeDescriptor.BlockTargetPair;
import org.apache.hadoop.hdfs.server.namenode.FSNamesystem;
import org.apache.hadoop.hdfs.server.protocol.DatanodeRegistration;
import org.apache.hadoop.hdfs.server.protocol.DatanodeStorage;
import org.apache.hadoop.net.NetworkTopology;
import org.junit.Before;
import org.junit.Test;
import org.mockito.Mockito;

import com.google.common.base.Joiner;
import com.google.common.collect.ImmutableList;
import com.google.common.collect.LinkedListMultimap;
import com.google.common.collect.Lists;

public class TestBlockManager {
  private DatanodeStorageInfo[] storages;
  private List<DatanodeDescriptor> nodes;
  private List<DatanodeDescriptor> rackA;
  private List<DatanodeDescriptor> rackB;

  /**
   * Some of these tests exercise code which has some randomness involved -
   * ie even if there's a bug, they may pass because the random node selection
   * chooses the correct result.
   *
   * Since they're true unit tests and run quickly, we loop them a number
   * of times trying to trigger the incorrect behavior.
   */
  private static final int NUM_TEST_ITERS = 30;
 
  private static final int BLOCK_SIZE = 64*1024;
 
  private Configuration conf;
  private FSNamesystem fsn;
  private BlockManager bm;

  @Before
  public void setupMockCluster() throws IOException {
    conf = new HdfsConfiguration();
    conf.set(DFSConfigKeys.NET_TOPOLOGY_SCRIPT_FILE_NAME_KEY,
        "need to set a dummy value here so it assumes a multi-rack cluster");
    fsn = Mockito.mock(FSNamesystem.class);
    Mockito.doReturn(true).when(fsn).hasWriteLock();
    bm = new BlockManager(fsn, fsn, conf);
    final String[] racks = {
        "/rackA",
        "/rackA",
        "/rackA",
        "/rackB",
        "/rackB",
        "/rackB"};
    storages = DFSTestUtil.createDatanodeStorageInfos(racks);
    nodes = Arrays.asList(DFSTestUtil.toDatanodeDescriptor(storages));
    rackA = nodes.subList(0, 3);
    rackB = nodes.subList(3, 6);
  }

  private void addNodes(Iterable<DatanodeDescriptor> nodesToAdd) {
    NetworkTopology cluster = bm.getDatanodeManager().getNetworkTopology();
    // construct network topology
    for (DatanodeDescriptor dn : nodesToAdd) {
      cluster.add(dn);
      dn.getStorageInfos()[0].setUtilizationForTesting(
          2 * HdfsConstants.MIN_BLOCKS_FOR_WRITE*BLOCK_SIZE, 0L,
          2 * HdfsConstants.MIN_BLOCKS_FOR_WRITE*BLOCK_SIZE, 0L);
      dn.updateHeartbeat(
          BlockManagerTestUtil.getStorageReportsForDatanode(dn), 0L, 0L, 0, 0);
      bm.getDatanodeManager().checkIfClusterIsNowMultiRack(dn);
    }
  }

  private void removeNode(DatanodeDescriptor deadNode) {
    NetworkTopology cluster = bm.getDatanodeManager().getNetworkTopology();
    cluster.remove(deadNode);
    bm.removeBlocksAssociatedTo(deadNode);
  }


  /**
   * Test that replication of under-replicated blocks is detected
   * and basically works
   */
  @Test
  public void testBasicReplication() throws Exception {
    addNodes(nodes);
    for (int i = 0; i < NUM_TEST_ITERS; i++) {
      doBasicTest(i);
    }
  }
 
  private void doBasicTest(int testIndex) {
    List<DatanodeStorageInfo> origStorages = getStorages(0, 1);
    List<DatanodeDescriptor> origNodes = getNodes(origStorages);
    BlockInfo blockInfo = addBlockOnNodes(testIndex, origNodes);

    DatanodeStorageInfo[] pipeline = scheduleSingleReplication(blockInfo);
    assertEquals(2, pipeline.length);
    assertTrue("Source of replication should be one of the nodes the block " +
        "was on. Was: " + pipeline[0],
        origStorages.contains(pipeline[0]));
    assertTrue("Destination of replication should be on the other rack. " +
        "Was: " + pipeline[1],
        rackB.contains(pipeline[1].getDatanodeDescriptor()));
  }
 

  /**
   * Regression test for HDFS-1480
   * - Cluster has 2 racks, A and B, each with three nodes.
   * - Block initially written on A1, A2, B1
   * - Admin decommissions two of these nodes (let's say A1 and A2 but it doesn't matter)
   * - Re-replication should respect rack policy
   */
  @Test
  public void testTwoOfThreeNodesDecommissioned() throws Exception {
    addNodes(nodes);
    for (int i = 0; i < NUM_TEST_ITERS; i++) {
      doTestTwoOfThreeNodesDecommissioned(i);
    }
  }
 
  private void doTestTwoOfThreeNodesDecommissioned(int testIndex) throws Exception {
    // Block originally on A1, A2, B1
    List<DatanodeStorageInfo> origStorages = getStorages(0, 1, 3);
    List<DatanodeDescriptor> origNodes = getNodes(origStorages);
    BlockInfo blockInfo = addBlockOnNodes(testIndex, origNodes);
   
    // Decommission two of the nodes (A1, A2)
    List<DatanodeDescriptor> decomNodes = startDecommission(0, 1);
   
    DatanodeStorageInfo[] pipeline = scheduleSingleReplication(blockInfo);
    assertTrue("Source of replication should be one of the nodes the block " +
        "was on. Was: " + pipeline[0],
        origStorages.contains(pipeline[0]));
    assertEquals("Should have three targets", 3, pipeline.length);
   
    boolean foundOneOnRackA = false;
    for (int i = 1; i < pipeline.length; i++) {
      DatanodeDescriptor target = pipeline[i].getDatanodeDescriptor();
      if (rackA.contains(target)) {
        foundOneOnRackA = true;
      }
      assertFalse(decomNodes.contains(target));
      assertFalse(origNodes.contains(target));
    }
   
    assertTrue("Should have at least one target on rack A. Pipeline: " +
        Joiner.on(",").join(pipeline),
        foundOneOnRackA);
  }
 

  /**
   * Test what happens when a block is on three nodes, and all three of those
   * nodes are decommissioned. It should properly re-replicate to three new
   * nodes.
   */
  @Test
  public void testAllNodesHoldingReplicasDecommissioned() throws Exception {
    addNodes(nodes);
    for (int i = 0; i < NUM_TEST_ITERS; i++) {
      doTestAllNodesHoldingReplicasDecommissioned(i);
    }
  }

  private void doTestAllNodesHoldingReplicasDecommissioned(int testIndex) throws Exception {
    // Block originally on A1, A2, B1
    List<DatanodeStorageInfo> origStorages = getStorages(0, 1, 3);
    List<DatanodeDescriptor> origNodes = getNodes(origStorages);
    BlockInfo blockInfo = addBlockOnNodes(testIndex, origNodes);
   
    // Decommission all of the nodes
    List<DatanodeDescriptor> decomNodes = startDecommission(0, 1, 3);
   
    DatanodeStorageInfo[] pipeline = scheduleSingleReplication(blockInfo);
    assertTrue("Source of replication should be one of the nodes the block " +
        "was on. Was: " + pipeline[0],
        origStorages.contains(pipeline[0]));
    assertEquals("Should have three targets", 4, pipeline.length);
   
    boolean foundOneOnRackA = false;
    boolean foundOneOnRackB = false;
    for (int i = 1; i < pipeline.length; i++) {
      DatanodeDescriptor target = pipeline[i].getDatanodeDescriptor();
      if (rackA.contains(target)) {
        foundOneOnRackA = true;
      } else if (rackB.contains(target)) {
        foundOneOnRackB = true;
      }
      assertFalse(decomNodes.contains(target));
      assertFalse(origNodes.contains(target));
    }
   
    assertTrue("Should have at least one target on rack A. Pipeline: " +
        Joiner.on(",").join(pipeline),
        foundOneOnRackA);
    assertTrue("Should have at least one target on rack B. Pipeline: " +
        Joiner.on(",").join(pipeline),
        foundOneOnRackB);
  }

  /**
   * Test what happens when there are two racks, and an entire rack is
   * decommissioned.
   *
   * Since the cluster is multi-rack, it will consider the block
   * under-replicated rather than create a third replica on the
   * same rack. Adding a new node on a third rack should cause re-replication
   * to that node.
   */
  @Test
  public void testOneOfTwoRacksDecommissioned() throws Exception {
    addNodes(nodes);
    for (int i = 0; i < NUM_TEST_ITERS; i++) {
      doTestOneOfTwoRacksDecommissioned(i);
    }
  }
 
  private void doTestOneOfTwoRacksDecommissioned(int testIndex) throws Exception {
    // Block originally on A1, A2, B1
    List<DatanodeStorageInfo> origStorages = getStorages(0, 1, 3);
    List<DatanodeDescriptor> origNodes = getNodes(origStorages);
    BlockInfo blockInfo = addBlockOnNodes(testIndex, origNodes);
   
    // Decommission all of the nodes in rack A
    List<DatanodeDescriptor> decomNodes = startDecommission(0, 1, 2);
   
    DatanodeStorageInfo[] pipeline = scheduleSingleReplication(blockInfo);
    assertTrue("Source of replication should be one of the nodes the block " +
        "was on. Was: " + pipeline[0],
        origStorages.contains(pipeline[0]));
    assertEquals("Should have three targets", 3, pipeline.length);
   
    boolean foundOneOnRackB = false;
    for (int i = 1; i < pipeline.length; i++) {
      DatanodeDescriptor target = pipeline[i].getDatanodeDescriptor();
      if (rackB.contains(target)) {
        foundOneOnRackB = true;
      }
      assertFalse(decomNodes.contains(target));
      assertFalse(origNodes.contains(target));
    }
   
    assertTrue("Should have at least one target on rack B. Pipeline: " +
        Joiner.on(",").join(pipeline),
        foundOneOnRackB);
   
    // Mark the block as received on the target nodes in the pipeline
    fulfillPipeline(blockInfo, pipeline);

    // the block is still under-replicated. Add a new node. This should allow
    // the third off-rack replica.
    DatanodeDescriptor rackCNode =
      DFSTestUtil.getDatanodeDescriptor("7.7.7.7", "/rackC");
    rackCNode.updateStorage(new DatanodeStorage(DatanodeStorage.generateUuid()));
    addNodes(ImmutableList.of(rackCNode));
    try {
      DatanodeStorageInfo[] pipeline2 = scheduleSingleReplication(blockInfo);
      assertEquals(2, pipeline2.length);
      assertEquals(rackCNode, pipeline2[1].getDatanodeDescriptor());
    } finally {
      removeNode(rackCNode);
    }
  }

  /**
   * Unit test version of testSufficientlyReplBlocksUsesNewRack from
   * {@link TestBlocksWithNotEnoughRacks}.
   **/
  @Test
  public void testSufficientlyReplBlocksUsesNewRack() throws Exception {
    addNodes(nodes);
    for (int i = 0; i < NUM_TEST_ITERS; i++) {
      doTestSufficientlyReplBlocksUsesNewRack(i);
    }
  }

  private void doTestSufficientlyReplBlocksUsesNewRack(int testIndex) {
    // Originally on only nodes in rack A.
    List<DatanodeDescriptor> origNodes = rackA;
    BlockInfo blockInfo = addBlockOnNodes(testIndex, origNodes);
    DatanodeStorageInfo pipeline[] = scheduleSingleReplication(blockInfo);
   
    assertEquals(2, pipeline.length); // single new copy
    assertTrue("Source of replication should be one of the nodes the block " +
        "was on. Was: " + pipeline[0],
        origNodes.contains(pipeline[0].getDatanodeDescriptor()));
    assertTrue("Destination of replication should be on the other rack. " +
        "Was: " + pipeline[1],
        rackB.contains(pipeline[1].getDatanodeDescriptor()));
  }
 
  @Test
  public void testBlocksAreNotUnderreplicatedInSingleRack() throws Exception {
    List<DatanodeDescriptor> nodes = ImmutableList.of(
        BlockManagerTestUtil.getDatanodeDescriptor("1.1.1.1", "/rackA", true),
        BlockManagerTestUtil.getDatanodeDescriptor("2.2.2.2", "/rackA", true),
        BlockManagerTestUtil.getDatanodeDescriptor("3.3.3.3", "/rackA", true),
        BlockManagerTestUtil.getDatanodeDescriptor("4.4.4.4", "/rackA", true),
        BlockManagerTestUtil.getDatanodeDescriptor("5.5.5.5", "/rackA", true),
        BlockManagerTestUtil.getDatanodeDescriptor("6.6.6.6", "/rackA", true)
      );
    addNodes(nodes);
    List<DatanodeDescriptor> origNodes = nodes.subList(0, 3);
    for (int i = 0; i < NUM_TEST_ITERS; i++) {
      doTestSingleRackClusterIsSufficientlyReplicated(i, origNodes);
    }
  }
 
  private void doTestSingleRackClusterIsSufficientlyReplicated(int testIndex,
      List<DatanodeDescriptor> origNodes)
      throws Exception {
    assertEquals(0, bm.numOfUnderReplicatedBlocks());
    addBlockOnNodes(testIndex, origNodes);
    bm.processMisReplicatedBlocks();
    assertEquals(0, bm.numOfUnderReplicatedBlocks());
  }
 
 
  /**
   * Tell the block manager that replication is completed for the given
   * pipeline.
   */
  private void fulfillPipeline(BlockInfo blockInfo,
      DatanodeStorageInfo[] pipeline) throws IOException {
    for (int i = 1; i < pipeline.length; i++) {
      DatanodeStorageInfo storage = pipeline[i];
      bm.addBlock(storage.getDatanodeDescriptor(), storage.getStorageID(), blockInfo, null);
      blockInfo.addStorage(storage);
    }
  }

  private BlockInfo blockOnNodes(long blkId, List<DatanodeDescriptor> nodes) {
    Block block = new Block(blkId);
    BlockInfo blockInfo = new BlockInfo(block, 3);

    for (DatanodeDescriptor dn : nodes) {
      for (DatanodeStorageInfo storage : dn.getStorageInfos()) {
        blockInfo.addStorage(storage);
      }
    }
    return blockInfo;
  }

  private List<DatanodeDescriptor> getNodes(int ... indexes) {
    List<DatanodeDescriptor> ret = Lists.newArrayList();
    for (int idx : indexes) {
      ret.add(nodes.get(idx));
    }
    return ret;
  }

  private List<DatanodeDescriptor> getNodes(List<DatanodeStorageInfo> storages) {
    List<DatanodeDescriptor> ret = Lists.newArrayList();
    for (DatanodeStorageInfo s : storages) {
      ret.add(s.getDatanodeDescriptor());
    }
    return ret;
  }

  private List<DatanodeStorageInfo> getStorages(int ... indexes) {
    List<DatanodeStorageInfo> ret = Lists.newArrayList();
    for (int idx : indexes) {
      ret.add(storages[idx]);
    }
    return ret;
  }
 
  private List<DatanodeDescriptor> startDecommission(int ... indexes) {
    List<DatanodeDescriptor> nodes = getNodes(indexes);
    for (DatanodeDescriptor node : nodes) {
      node.startDecommission();
    }
    return nodes;
  }
 
  private BlockInfo addBlockOnNodes(long blockId, List<DatanodeDescriptor> nodes) {
    BlockCollection bc = Mockito.mock(BlockCollection.class);
    Mockito.doReturn((short)3).when(bc).getBlockReplication();
    BlockInfo blockInfo = blockOnNodes(blockId, nodes);

    bm.blocksMap.addBlockCollection(blockInfo, bc);
    return blockInfo;
  }

  private DatanodeStorageInfo[] scheduleSingleReplication(Block block) {
    // list for priority 1
    List<Block> list_p1 = new ArrayList<Block>();
    list_p1.add(block);

    // list of lists for each priority
    List<List<Block>> list_all = new ArrayList<List<Block>>();
    list_all.add(new ArrayList<Block>()); // for priority 0
    list_all.add(list_p1); // for priority 1

    assertEquals("Block not initially pending replication", 0,
        bm.pendingReplications.getNumReplicas(block));
    assertEquals(
        "computeReplicationWork should indicate replication is needed", 1,
        bm.computeReplicationWorkForBlocks(list_all));
    assertTrue("replication is pending after work is computed",
        bm.pendingReplications.getNumReplicas(block) > 0);

    LinkedListMultimap<DatanodeStorageInfo, BlockTargetPair> repls = getAllPendingReplications();
    assertEquals(1, repls.size());
    Entry<DatanodeStorageInfo, BlockTargetPair> repl =
      repls.entries().iterator().next();
       
    DatanodeStorageInfo[] targets = repl.getValue().targets;

    DatanodeStorageInfo[] pipeline = new DatanodeStorageInfo[1 + targets.length];
    pipeline[0] = repl.getKey();
    System.arraycopy(targets, 0, pipeline, 1, targets.length);

    return pipeline;
  }

  private LinkedListMultimap<DatanodeStorageInfo, BlockTargetPair> getAllPendingReplications() {
    LinkedListMultimap<DatanodeStorageInfo, BlockTargetPair> repls =
      LinkedListMultimap.create();
    for (DatanodeDescriptor dn : nodes) {
      List<BlockTargetPair> thisRepls = dn.getReplicationCommand(10);
      if (thisRepls != null) {
        for(DatanodeStorageInfo storage : dn.getStorageInfos()) {
          repls.putAll(storage, thisRepls);
        }
      }
    }
    return repls;
  }

  /**
   * Test that a source node for a highest-priority replication is chosen even if all available
   * source nodes have reached their replication limits.
   */
  @Test
  public void testHighestPriReplSrcChosenDespiteMaxReplLimit() throws Exception {
    bm.maxReplicationStreams = 0;
    bm.replicationStreamsHardLimit = 1;

    long blockId = 42;         // arbitrary
    Block aBlock = new Block(blockId, 0, 0);

    List<DatanodeDescriptor> origNodes = getNodes(0, 1);
    // Add the block to the first node.
    addBlockOnNodes(blockId,origNodes.subList(0,1));

    List<DatanodeDescriptor> cntNodes = new LinkedList<DatanodeDescriptor>();
    List<DatanodeStorageInfo> liveNodes = new LinkedList<DatanodeStorageInfo>();

    assertNotNull("Chooses source node for a highest-priority replication"
        + " even if all available source nodes have reached their replication"
        + " limits below the hard limit.",
        bm.chooseSourceDatanode(
            aBlock,
            cntNodes,
            liveNodes,
            new NumberReplicas(),
            UnderReplicatedBlocks.QUEUE_HIGHEST_PRIORITY));

    assertNull("Does not choose a source node for a less-than-highest-priority"
        + " replication since all available source nodes have reached"
        + " their replication limits.",
        bm.chooseSourceDatanode(
            aBlock,
            cntNodes,
            liveNodes,
            new NumberReplicas(),
            UnderReplicatedBlocks.QUEUE_VERY_UNDER_REPLICATED));

    // Increase the replication count to test replication count > hard limit
    DatanodeStorageInfo targets[] = { origNodes.get(1).getStorageInfos()[0] };
    origNodes.get(0).addBlockToBeReplicated(aBlock, targets);

    assertNull("Does not choose a source node for a highest-priority"
        + " replication when all available nodes exceed the hard limit.",
        bm.chooseSourceDatanode(
            aBlock,
            cntNodes,
            liveNodes,
            new NumberReplicas(),
            UnderReplicatedBlocks.QUEUE_HIGHEST_PRIORITY));
  }

  @Test
  public void testSafeModeIBR() throws Exception {
    DatanodeDescriptor node = spy(nodes.get(0));
    DatanodeStorageInfo ds = node.getStorageInfos()[0];

    // TODO: Needs to be fixed. DatanodeUuid is not storageID.
    node.setDatanodeUuidForTesting(ds.getStorageID());

    node.isAlive = true;

    DatanodeRegistration nodeReg =
        new DatanodeRegistration(node, null, null, "");

    // pretend to be in safemode
    doReturn(true).when(fsn).isInStartupSafeMode();
   
    // register new node
    bm.getDatanodeManager().registerDatanode(nodeReg);
    bm.getDatanodeManager().addDatanode(node); // swap in spy   
    assertEquals(node, bm.getDatanodeManager().getDatanode(node));
    assertEquals(0, ds.getBlockReportCount());
    // send block report, should be processed
    reset(node);
   
    bm.processReport(node, new DatanodeStorage(ds.getStorageID()), "pool",
        new BlockListAsLongs(null, null));
    assertEquals(1, ds.getBlockReportCount());
    // send block report again, should NOT be processed
    reset(node);
    bm.processReport(node, new DatanodeStorage(ds.getStorageID()), "pool",
        new BlockListAsLongs(null, null));
    assertEquals(1, ds.getBlockReportCount());

    // re-register as if node restarted, should update existing node
    bm.getDatanodeManager().removeDatanode(node);
    reset(node);
    bm.getDatanodeManager().registerDatanode(nodeReg);
    verify(node).updateRegInfo(nodeReg);
    assertEquals(0, ds.getBlockReportCount()); // ready for report again
    // send block report, should be processed after restart
    reset(node);
    bm.processReport(node, new DatanodeStorage(ds.getStorageID()), "pool",
        new BlockListAsLongs(null, null));
    assertEquals(1, ds.getBlockReportCount());
  }
 
  @Test
  public void testSafeModeIBRAfterIncremental() throws Exception {
    DatanodeDescriptor node = spy(nodes.get(0));
    DatanodeStorageInfo ds = node.getStorageInfos()[0];

    // TODO: Needs to be fixed. DatanodeUuid is not storageID.
    node.setDatanodeUuidForTesting(ds.getStorageID());

    node.isAlive = true;

    DatanodeRegistration nodeReg =
        new DatanodeRegistration(node, null, null, "");

    // pretend to be in safemode
    doReturn(true).when(fsn).isInStartupSafeMode();

    // register new node
    bm.getDatanodeManager().registerDatanode(nodeReg);
    bm.getDatanodeManager().addDatanode(node); // swap in spy   
    assertEquals(node, bm.getDatanodeManager().getDatanode(node));
    assertEquals(0, ds.getBlockReportCount());
    // send block report while pretending to already have blocks
    reset(node);
    doReturn(1).when(node).numBlocks();
    bm.processReport(node, new DatanodeStorage(ds.getStorageID()), "pool",
        new BlockListAsLongs(null, null));
    assertEquals(1, ds.getBlockReportCount());
  }
}
TOP

Related Classes of org.apache.hadoop.hdfs.server.blockmanagement.TestBlockManager

TOP
Copyright © 2018 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.