HDFS-16314. Support to make dfs.namenode.block-placement-policy.exclude-slow-nodes.enabled reconfigurable (#3664)

Reviewed-by: Fei Hui <feihui.ustc@gmail.com>
Signed-off-by: Akira Ajisaka <aajisaka@apache.org>
(cherry picked from commit c0f405a46b402b4bdf5b7771f5124790fc866965)

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockManager.java

@@ -5319,4 +5319,14 @@ public class BlockManager implements BlockStatsMXBean {
   public StoragePolicySatisfyManager getSPSManager() {
     return spsManager;
   }
+
+  public void setExcludeSlowNodesEnabled(boolean enable) {
+    placementPolicies.getPolicy(CONTIGUOUS).setExcludeSlowNodesEnabled(enable);
+    placementPolicies.getPolicy(STRIPED).setExcludeSlowNodesEnabled(enable);
+  }
+
+  @VisibleForTesting
+  public boolean getExcludeSlowNodesEnabled(BlockType blockType) {
+    return placementPolicies.getPolicy(blockType).getExcludeSlowNodesEnabled();
+  }
 }

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockPlacementPolicy.java

@@ -261,4 +261,16 @@ public abstract class BlockPlacementPolicy {
       }
     }
   }
+
+  /**
+   * Updates the value used for excludeSlowNodesEnabled, which is set by
+   * {@code DFSConfigKeys.DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_KEY}
+   * initially.
+   *
+   * @param enable true, we will filter out slow nodes
+   * when choosing targets for blocks, otherwise false not filter.
+   */
+  public abstract void setExcludeSlowNodesEnabled(boolean enable);
+
+  public abstract boolean getExcludeSlowNodesEnabled();
 }

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/server/blockmanagement/BlockPlacementPolicyDefault.java

@@ -103,7 +103,7 @@ public class BlockPlacementPolicyDefault extends BlockPlacementPolicy {
   protected double considerLoadFactor;
   private boolean preferLocalNode;
   private boolean dataNodePeerStatsEnabled;
-  private boolean excludeSlowNodesEnabled;
+  private volatile boolean excludeSlowNodesEnabled;
   protected NetworkTopology clusterMap;
   protected Host2NodesMap host2datanodeMap;
   private FSClusterStats stats;
@@ -1359,5 +1359,15 @@ public class BlockPlacementPolicyDefault extends BlockPlacementPolicy {
   void setPreferLocalNode(boolean prefer) {
     this.preferLocalNode = prefer;
   }
+
+  @Override
+  public void setExcludeSlowNodesEnabled(boolean enable) {
+    this.excludeSlowNodesEnabled = enable;
+  }
+
+  @Override
+  public boolean getExcludeSlowNodesEnabled() {
+    return excludeSlowNodesEnabled;
+  }
 }
 

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

@@ -188,6 +188,8 @@ import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_BLOCK_REPLICATOR_CLASSNAM
 import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_BLOCK_PLACEMENT_EC_CLASSNAME_KEY;
 import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_KEY;
 import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_DEFAULT;
+import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_KEY;
+import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_DEFAULT;
 
 import static org.apache.hadoop.util.ExitUtil.terminate;
 import static org.apache.hadoop.util.ToolRunner.confirmPrompt;
@@ -330,7 +332,8 @@ public class NameNode extends ReconfigurableBase implements
           DFS_BLOCK_REPLICATOR_CLASSNAME_KEY,
           DFS_BLOCK_PLACEMENT_EC_CLASSNAME_KEY,
           DFS_IMAGE_PARALLEL_LOAD_KEY,
-          DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_KEY));
+          DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_KEY,
+          DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_KEY));
 
   private static final String USAGE = "Usage: hdfs namenode ["
       + StartupOption.BACKUP.getName() + "] | \n\t["
@@ -2196,7 +2199,8 @@ public class NameNode extends ReconfigurableBase implements
       return newVal;
     } else if (property.equals(DFS_IMAGE_PARALLEL_LOAD_KEY)) {
       return reconfigureParallelLoad(newVal);
-    } else if (property.equals(DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_KEY)) {
+    } else if (property.equals(DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_KEY)
+          || (property.equals(DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_KEY))) {
       return reconfigureSlowNodesParameters(datanodeManager, property, newVal);
     } else {
       throw new ReconfigurationException(property, newVal, getConf().get(
@@ -2230,7 +2234,7 @@ public class NameNode extends ReconfigurableBase implements
         newSetting = bm.getBlocksReplWorkMultiplier();
       } else {
         throw new IllegalArgumentException("Unexpected property " +
-            property + "in reconfReplicationParameters");
+            property + " in reconfReplicationParameters");
       }
       LOG.info("RECONFIGURE* changed {} to {}", property, newSetting);
       return String.valueOf(newSetting);
@@ -2386,15 +2390,24 @@ public class NameNode extends ReconfigurableBase implements
 
   String reconfigureSlowNodesParameters(final DatanodeManager datanodeManager,
       final String property, final String newVal) throws ReconfigurationException {
+    BlockManager bm = namesystem.getBlockManager();
     namesystem.writeLock();
     boolean enable;
     try {
-      if (newVal == null) {
-        enable = DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_DEFAULT;
+      if (property.equals(DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_KEY)) {
+        enable = (newVal == null ? DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_DEFAULT :
+            Boolean.parseBoolean(newVal));
+        datanodeManager.setAvoidSlowDataNodesForReadEnabled(enable);
+      } else if (property.equals(
+            DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_KEY)) {
+        enable = (newVal == null ?
+            DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_DEFAULT :
+            Boolean.parseBoolean(newVal));
+        bm.setExcludeSlowNodesEnabled(enable);
       } else {
-        enable = Boolean.parseBoolean(newVal);
+        throw new IllegalArgumentException("Unexpected property " +
+            property + " in reconfigureSlowNodesParameters");
       }
-      datanodeManager.setAvoidSlowDataNodesForReadEnabled(enable);
       LOG.info("RECONFIGURE* changed {} to {}", property, newVal);
       return Boolean.toString(enable);
     } catch (IllegalArgumentException e) {

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

@@ -34,10 +34,12 @@ import org.apache.hadoop.conf.ReconfigurationException;
 import org.apache.hadoop.fs.Path;
 import org.apache.hadoop.hdfs.MiniDFSCluster;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants.StoragePolicySatisfierMode;
+import org.apache.hadoop.hdfs.protocol.BlockType;
 import org.apache.hadoop.hdfs.DFSConfigKeys;
 import org.apache.hadoop.hdfs.DistributedFileSystem;
 import org.apache.hadoop.hdfs.HdfsConfiguration;
 import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeManager;
+import org.apache.hadoop.hdfs.server.blockmanagement.BlockManager;
 import org.apache.hadoop.hdfs.server.namenode.sps.StoragePolicySatisfyManager;
 import org.apache.hadoop.ipc.RemoteException;
 import org.apache.hadoop.test.GenericTestUtils;
@@ -52,6 +54,7 @@ import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_STORAGE_POLICY_SATISFIER_
 import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_STORAGE_POLICY_SATISFIER_MODE_DEFAULT;
 import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_BLOCK_INVALIDATE_LIMIT_KEY;
 import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_KEY;
+import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_KEY;
 import static org.apache.hadoop.fs.CommonConfigurationKeys.IPC_BACKOFF_ENABLE_DEFAULT;
 
 public class TestNameNodeReconfigure {
@@ -399,8 +402,8 @@ public class TestNameNodeReconfigure {
   public void testEnableSlowNodesParametersAfterReconfigured()
       throws ReconfigurationException {
     final NameNode nameNode = cluster.getNameNode();
-    final DatanodeManager datanodeManager = nameNode.namesystem
-        .getBlockManager().getDatanodeManager();
+    final BlockManager blockManager = nameNode.namesystem.getBlockManager();
+    final DatanodeManager datanodeManager = blockManager.getDatanodeManager();
 
     // By default, avoidSlowDataNodesForRead is false.
     assertEquals(false, datanodeManager.getEnableAvoidSlowDataNodesForRead());
@@ -410,6 +413,21 @@ public class TestNameNodeReconfigure {
 
     // After reconfigured, avoidSlowDataNodesForRead is true.
     assertEquals(true, datanodeManager.getEnableAvoidSlowDataNodesForRead());
+
+    // By default, excludeSlowNodesEnabled is false.
+    assertEquals(false, blockManager.
+        getExcludeSlowNodesEnabled(BlockType.CONTIGUOUS));
+    assertEquals(false, blockManager.
+        getExcludeSlowNodesEnabled(BlockType.STRIPED));
+
+    nameNode.reconfigureProperty(
+        DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_KEY, Boolean.toString(true));
+
+    // After reconfigured, excludeSlowNodesEnabled is true.
+    assertEquals(true, blockManager.
+        getExcludeSlowNodesEnabled(BlockType.CONTIGUOUS));
+    assertEquals(true, blockManager.
+        getExcludeSlowNodesEnabled(BlockType.STRIPED));
   }
 
   @After

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/tools/TestDFSAdmin.java

@@ -26,6 +26,7 @@ import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_HEARTBEAT_RECHEC
 import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_BLOCK_PLACEMENT_EC_CLASSNAME_KEY;
 import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_BLOCK_REPLICATOR_CLASSNAME_KEY;
 import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_KEY;
+import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_KEY;
 
 import java.util.function.Supplier;
 import org.apache.hadoop.thirdparty.com.google.common.collect.Lists;
@@ -423,13 +424,14 @@ public class TestDFSAdmin {
     final List<String> outs = Lists.newArrayList();
     final List<String> errs = Lists.newArrayList();
     getReconfigurableProperties("namenode", address, outs, errs);
-    assertEquals(14, outs.size());
+    assertEquals(15, outs.size());
     assertEquals(DFS_BLOCK_PLACEMENT_EC_CLASSNAME_KEY, outs.get(1));
     assertEquals(DFS_BLOCK_REPLICATOR_CLASSNAME_KEY, outs.get(2));
     assertEquals(DFS_HEARTBEAT_INTERVAL_KEY, outs.get(3));
     assertEquals(DFS_IMAGE_PARALLEL_LOAD_KEY, outs.get(4));
     assertEquals(DFS_NAMENODE_AVOID_SLOW_DATANODE_FOR_READ_KEY, outs.get(5));
-    assertEquals(DFS_NAMENODE_HEARTBEAT_RECHECK_INTERVAL_KEY, outs.get(6));
+    assertEquals(DFS_NAMENODE_BLOCKPLACEMENTPOLICY_EXCLUDE_SLOW_NODES_ENABLED_KEY, outs.get(6));
+    assertEquals(DFS_NAMENODE_HEARTBEAT_RECHECK_INTERVAL_KEY, outs.get(7));
     assertEquals(errs.size(), 0);
   }