HDFS-17223. Add journalnode maintenance node list

Author: 顾鹏

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSConfigKeys.java

@@ -1466,6 +1466,10 @@ public class DFSConfigKeys extends CommonConfigurationKeys {
           "dfs.journalnode.edit-cache-size.fraction";
   public static final float DFS_JOURNALNODE_EDIT_CACHE_SIZE_FRACTION_DEFAULT = 0.5f;
 
+  public static final String  DFS_JOURNALNODE_MAINTENANCE_NODES_KEY =
+      "dfs.journalnode.maintenance.nodes";
+  public static final String[]  DFS_JOURNALNODE_MAINTENANCE_NODES_DEFAULT = {};
+
   // Journal-node related configs for the client side.
   public static final String  DFS_QJOURNAL_QUEUE_SIZE_LIMIT_KEY = "dfs.qjournal.queued-edits.limit.mb";
   public static final int     DFS_QJOURNAL_QUEUE_SIZE_LIMIT_DEFAULT = 10;

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/DFSUtil.java

@@ -70,6 +70,7 @@
 import org.apache.hadoop.classification.InterfaceStability;
 import org.apache.hadoop.fs.ParentNotDirectoryException;
 import org.apache.hadoop.fs.UnresolvedLinkException;
+import org.apache.hadoop.hdfs.server.blockmanagement.HostSet;
 import org.apache.hadoop.hdfs.server.datanode.metrics.DataNodeMetrics;
 import org.apache.hadoop.hdfs.server.namenode.FSDirectory;
 import org.apache.hadoop.hdfs.server.namenode.INodesInPath;
@@ -1982,4 +1983,32 @@ public static void addTransferRateMetric(final DataNodeMetrics metrics, final lo
       LOG.warn("Unexpected value for data transfer bytes={} duration={}", read, duration);
     }
   }
+
+  /**
+   * Construct a HostSet from an array of "ip:port" strings.
+   * @param nodesHostPort ip port string array.
+   * @return HostSet of InetSocketAddress.
+   */
+  public static HostSet getHostSet(String[] nodesHostPort) {
+    HostSet retSet = new HostSet();
+    for (String hostPort : nodesHostPort) {
+      try {
+        URI uri = new URI("dummy", hostPort, null, null, null);
+        int port = uri.getPort();
+        if (port < 0) {
+          LOG.warn(String.format("The ip:port `%s` is invalid, skip this node.", hostPort));
+          continue;
+        }
+        InetSocketAddress inetSocketAddress = new InetSocketAddress(uri.getHost(), port);
+        if (inetSocketAddress.isUnresolved()) {
+          LOG.warn(String.format("Failed to resolve address `%s`", hostPort));
+          continue;
+        }
+        retSet.add(inetSocketAddress);
+      } catch (URISyntaxException e) {
+        LOG.warn(String.format("Failed to parse `%s`", hostPort));
+      }
+    }
+    return retSet;
+  }
 }

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/qjournal/client/AsyncLoggerSet.java

@@ -53,9 +53,15 @@ class AsyncLoggerSet {
 
   private static final long INVALID_EPOCH = -1;
   private long myEpoch = INVALID_EPOCH;
+  private final int majoritySize;
 
-  public AsyncLoggerSet(List<AsyncLogger> loggers) {
+  AsyncLoggerSet(List<AsyncLogger> loggers) {
+    this(loggers, loggers.size());
+  }
+
+  AsyncLoggerSet(List<AsyncLogger> loggers, int quorumJournalCount) {
     this.loggers = ImmutableList.copyOf(loggers);
+    this.majoritySize = quorumJournalCount / 2 + 1;
   }
 
   void setEpoch(long e) {
@@ -151,7 +157,7 @@ <V> Map<AsyncLogger, V> waitForWriteQuorum(QuorumCall<AsyncLogger, V> q,
    * @return the number of nodes which are required to obtain a quorum.
    */
   int getMajoritySize() {
-    return loggers.size() / 2 + 1;
+    return this.majoritySize;
   }
 
   /**
@@ -161,6 +167,11 @@ String getMajorityString() {
     return getMajoritySize() + "/" + loggers.size();
   }
 
+  @VisibleForTesting
+  List<AsyncLogger> getLoggerListForTests() {
+    return loggers;
+  }
+
   /**
    * @return the number of loggers behind this set
    */

hadoop-hdfs-project/hadoop-hdfs/src/main/java/org/apache/hadoop/hdfs/qjournal/client/QuorumJournalManager.java

@@ -17,6 +17,9 @@
  */
 package org.apache.hadoop.hdfs.qjournal.client;
 
+import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_JOURNALNODE_MAINTENANCE_NODES_DEFAULT;
+import static org.apache.hadoop.hdfs.DFSConfigKeys.DFS_JOURNALNODE_MAINTENANCE_NODES_KEY;
+
 import java.io.IOException;
 import java.net.InetSocketAddress;
 import java.net.URI;
@@ -31,6 +34,7 @@
 import java.util.concurrent.TimeUnit;
 import java.util.concurrent.TimeoutException;
 
+import org.apache.hadoop.hdfs.server.blockmanagement.HostSet;
 import org.apache.hadoop.util.Lists;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
@@ -108,6 +112,7 @@ public class QuorumJournalManager implements JournalManager {
   private static final int OUTPUT_BUFFER_CAPACITY_DEFAULT = 512 * 1024;
   private int outputBufferCapacity;
   private final URLConnectionFactory connectionFactory;
+  private int quorumJournalCount;
 
   /** Limit logging about input stream selection to every 5 seconds max. */
   private static final long SELECT_INPUT_STREAM_LOG_INTERVAL_MS = 5000;
@@ -144,7 +149,18 @@ public QuorumJournalManager(Configuration conf,
     this.uri = uri;
     this.nsInfo = nsInfo;
     this.nameServiceId = nameServiceId;
-    this.loggers = new AsyncLoggerSet(createLoggers(loggerFactory));
+
+    // createLoggers() will set quorumJournalCount to total number of journal nodes while return a
+    // list of healthy/good journal nodes.
+    List<AsyncLogger> asyncLoggerList = createLoggers(loggerFactory);
+    this.loggers = new AsyncLoggerSet(asyncLoggerList, this.quorumJournalCount);
+
+    // Check whether the number of jn maintenance lists is valid
+    int quorumThreshold = quorumJournalCount / 2 + 1;
+    Preconditions.checkArgument(
+        this.loggers.size() >= quorumThreshold,
+        "The total journalnode minus %s the number of blacklists must be greater than or equal to"
+            + " %s!", DFS_JOURNALNODE_MAINTENANCE_NODES_KEY, quorumThreshold);
 
     this.maxTxnsPerRpc =
         conf.getInt(QJM_RPC_MAX_TXNS_KEY, QJM_RPC_MAX_TXNS_DEFAULT);
@@ -250,6 +266,10 @@ Map<AsyncLogger, NewEpochResponseProto> createNewUniqueEpoch()
 
   @Override
   public void format(NamespaceInfo nsInfo, boolean force) throws IOException {
+    if (isJNInMaintenanceMode()) {
+      throw new IOException(
+          "Formatting a journal node is not support while in jn maintenance mode");
+    }
     QuorumCall<AsyncLogger, Void> call = loggers.format(nsInfo, force);
     try {
       call.waitFor(loggers.size(), loggers.size(), 0, timeoutMs,
@@ -406,21 +426,39 @@ private void recoverUnclosedSegment(long segmentTxId) throws IOException {
             logToSync.getStartTxId(),
             logToSync.getEndTxId()));
   }
-  
-  static List<AsyncLogger> createLoggers(Configuration conf,
+
+  List<AsyncLogger> createLoggers(Configuration conf,
+                                  URI uri,
+                                  NamespaceInfo nsInfo,
+                                  AsyncLogger.Factory factory,
+                                  String nameServiceId)
+      throws IOException {
+    String[] skipNodesHostPort = conf.getTrimmedStrings(
+        DFS_JOURNALNODE_MAINTENANCE_NODES_KEY, DFS_JOURNALNODE_MAINTENANCE_NODES_DEFAULT);
+    return createLoggers(conf, uri, nsInfo, factory, nameServiceId, skipNodesHostPort);
+  }
+
+  private List<AsyncLogger> createLoggers(Configuration conf,
                                          URI uri,
                                          NamespaceInfo nsInfo,
                                          AsyncLogger.Factory factory,
-                                         String nameServiceId)
+                                         String nameServiceId,
+                                         String[] skipNodesHostPort)
       throws IOException {
     List<AsyncLogger> ret = Lists.newArrayList();
     List<InetSocketAddress> addrs = Util.getAddressesList(uri, conf);
     if (addrs.size() % 2 == 0) {
       LOG.warn("Quorum journal URI '" + uri + "' has an even number " +
           "of Journal Nodes specified. This is not recommended!");
     }
+    setQuorumJournalCount(addrs.size());
+    HostSet skipSet = DFSUtil.getHostSet(skipNodesHostPort);
     String jid = parseJournalId(uri);
     for (InetSocketAddress addr : addrs) {
+      if(skipSet.match(addr)) {
+        LOG.info("The node {} is a maintenance node and will be skipped.", addr);
+        continue;
+      }
       ret.add(factory.createLogger(conf, nsInfo, jid, nameServiceId, addr));
     }
     return ret;
@@ -667,6 +705,9 @@ AsyncLoggerSet getLoggerSetForTests() {
 
   @Override
   public void doPreUpgrade() throws IOException {
+    if (isJNInMaintenanceMode()) {
+      throw new IOException("doPreUpgrade() is not support while in jn maintenance mode");
+    }
     QuorumCall<AsyncLogger, Void> call = loggers.doPreUpgrade();
     try {
       call.waitFor(loggers.size(), loggers.size(), 0, timeoutMs,
@@ -684,6 +725,9 @@ public void doPreUpgrade() throws IOException {
 
   @Override
   public void doUpgrade(Storage storage) throws IOException {
+    if (isJNInMaintenanceMode()) {
+      throw new IOException("doUpgrade() is not support while in jn maintenance mode");
+    }
     QuorumCall<AsyncLogger, Void> call = loggers.doUpgrade(storage);
     try {
       call.waitFor(loggers.size(), loggers.size(), 0, timeoutMs,
@@ -701,6 +745,9 @@ public void doUpgrade(Storage storage) throws IOException {
 
   @Override
   public void doFinalize() throws IOException {
+    if (isJNInMaintenanceMode()) {
+      throw new IOException("doFinalize() is not support while in jn maintenance mode");
+    }
     QuorumCall<AsyncLogger, Void> call = loggers.doFinalize();
     try {
       call.waitFor(loggers.size(), loggers.size(), 0, timeoutMs,
@@ -719,6 +766,9 @@ public void doFinalize() throws IOException {
   @Override
   public boolean canRollBack(StorageInfo storage, StorageInfo prevStorage,
       int targetLayoutVersion) throws IOException {
+    if (isJNInMaintenanceMode()) {
+      throw new IOException("canRollBack() is not support while in jn maintenance mode");
+    }
     QuorumCall<AsyncLogger, Boolean> call = loggers.canRollBack(storage,
         prevStorage, targetLayoutVersion);
     try {
@@ -753,6 +803,9 @@ public boolean canRollBack(StorageInfo storage, StorageInfo prevStorage,
 
   @Override
   public void doRollback() throws IOException {
+    if (isJNInMaintenanceMode()) {
+      throw new IOException("doRollback() is not support while in jn maintenance mode");
+    }
     QuorumCall<AsyncLogger, Void> call = loggers.doRollback();
     try {
       call.waitFor(loggers.size(), loggers.size(), 0, timeoutMs,
@@ -770,6 +823,9 @@ public void doRollback() throws IOException {
 
   @Override
   public void discardSegments(long startTxId) throws IOException {
+    if (isJNInMaintenanceMode()) {
+      throw new IOException("discardSegments() is not support while in jn maintenance mode");
+    }
     QuorumCall<AsyncLogger, Void> call = loggers.discardSegments(startTxId);
     try {
       call.waitFor(loggers.size(), loggers.size(), 0,
@@ -789,6 +845,9 @@ public void discardSegments(long startTxId) throws IOException {
 
   @Override
   public long getJournalCTime() throws IOException {
+    if (isJNInMaintenanceMode()) {
+      throw new IOException("getJournalCTime() is not support while in jn maintenance mode");
+    }
     QuorumCall<AsyncLogger, Long> call = loggers.getJournalCTime();
     try {
       call.waitFor(loggers.size(), loggers.size(), 0,
@@ -819,4 +878,12 @@ public long getJournalCTime() throws IOException {
 
     throw new AssertionError("Unreachable code.");
   }
+
+  public void setQuorumJournalCount(int quorumJournalCount) {
+    this.quorumJournalCount = quorumJournalCount;
+  }
+
+  private boolean isJNInMaintenanceMode() {
+    return this.loggers.size() < quorumJournalCount;
+  }
 }

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

@@ -45,7 +45,7 @@ public class HostSet implements Iterable<InetSocketAddress> {
    * The function that checks whether there exists an entry foo in the set
    * so that foo &lt;= addr.
    */
-  boolean matchedBy(InetSocketAddress addr) {
+  public boolean matchedBy(InetSocketAddress addr) {
     Collection<Integer> ports = addrs.get(addr.getAddress());
     return addr.getPort() == 0 ? !ports.isEmpty() : ports.contains(addr
         .getPort());
@@ -55,23 +55,23 @@ boolean matchedBy(InetSocketAddress addr) {
    * The function that checks whether there exists an entry foo in the set
    * so that addr &lt;= foo.
    */
-  boolean match(InetSocketAddress addr) {
+  public boolean match(InetSocketAddress addr) {
     int port = addr.getPort();
     Collection<Integer> ports = addrs.get(addr.getAddress());
     boolean exactMatch = ports.contains(port);
     boolean genericMatch = ports.contains(0);
     return exactMatch || genericMatch;
   }
 
-  boolean isEmpty() {
+  public boolean isEmpty() {
     return addrs.isEmpty();
   }
 
-  int size() {
+  public int size() {
     return addrs.size();
   }
 
-  void add(InetSocketAddress addr) {
+  public void add(InetSocketAddress addr) {
     Preconditions.checkArgument(!addr.isUnresolved());
     addrs.put(addr.getAddress(), addr.getPort());
   }

hadoop-hdfs-project/hadoop-hdfs/src/main/resources/hdfs-default.xml

@@ -6333,6 +6333,23 @@
     </description>
   </property>
 
+  <property>
+    <name>dfs.journalnode.maintenance.nodes</name>
+    <value></value>
+    <description>
+      In the case that one out of three journal nodes being down, theoretically HDFS can still
+      function. However, in reality, the unavailable journal node may not recover quickly. During
+      this period, when we need to restart an Namenode, the Namenode will try to connect to the
+      unavailable journal node through the lengthy RPC retry mechanism, resulting in a long
+      initialization time for the Namenode. By adding these unavailable journal nodes to the
+      maintenance nodes, we will skip these unavailable journal nodes during Namenode initialization
+      and thus reduce namenode startup time.
+      1-node example values: jn01:8485
+      2-node example values: jn01:8485,jn02:8485
+    </description>
+  </property>
+
+
   <property>
     <name>dfs.namenode.lease-hard-limit-sec</name>
     <value>1200</value>

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

@@ -41,6 +41,7 @@
 import static org.junit.Assert.assertArrayEquals;
 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.assertThat;
 import static org.junit.Assert.assertTrue;
@@ -72,6 +73,7 @@
 import org.apache.hadoop.hdfs.protocol.HdfsConstants;
 import org.apache.hadoop.hdfs.protocol.LocatedBlock;
 import org.apache.hadoop.hdfs.protocol.LocatedBlocks;
+import org.apache.hadoop.hdfs.server.blockmanagement.HostSet;
 import org.apache.hadoop.hdfs.server.datanode.metrics.DataNodeMetrics;
 import org.apache.hadoop.hdfs.server.namenode.NameNode;
 import org.apache.hadoop.hdfs.server.namenode.ha.ConfiguredFailoverProxyProvider;
@@ -1137,4 +1139,24 @@ public void testAddTransferRateMetricForInvalidValue() {
     DFSUtil.addTransferRateMetric(mockMetrics, 100, 0);
     verify(mockMetrics, times(0)).addReadTransferRate(anyLong());
   }
+
+  @Test
+  public void testGetHostSet() {
+    String[] testAddrs = new String[] {"unreachable-host1.com:9000", "unreachable-host2.com:9000"};
+    HostSet hostSet = DFSUtil.getHostSet(testAddrs);
+    assertNotNull(hostSet);
+    assertEquals(0, hostSet.size());
+
+    String strAddress = "localhost";
+    testAddrs = new String[] {strAddress};
+    hostSet = DFSUtil.getHostSet(testAddrs);
+    assertEquals(0, hostSet.size());
+
+    strAddress = "localhost:9000";
+    InetSocketAddress inetSocketAddress = new InetSocketAddress("localhost", 9000);
+    testAddrs = new String[] {strAddress};
+    hostSet = DFSUtil.getHostSet(testAddrs);
+    assertNotNull(hostSet);
+    assertTrue(hostSet.match(inetSocketAddress));
+  }
 }