HDFS-8883. NameNode Metrics : Add FSNameSystem lock Queue Length. Contributed by Anu Engineer.

(cherry picked from commit a7862d5fe4)

hadoop-common-project/hadoop-common/src/site/markdown/Metrics.md

@@ -235,6 +235,7 @@ Each metrics record contains tags such as HAState and Hostname as additional inf
 | `NumActiveClients` | Current number of active clients holding lease |
 | `HAState` | (HA-only) Current state of the NameNode: initializing or active or standby or stopping state |
 | `FSState` | Current state of the file system: Safemode or Operational |
+| `LockQueueLength` | Number of threads waiting to acquire FSNameSystem lock |
 
 JournalNode
 -----------

hadoop-hdfs-project/hadoop-hdfs/CHANGES.txt

@@ -445,6 +445,9 @@ Release 2.8.0 - UNRELEASED
 
     HDFS-8824. Do not use small blocks for balancing the cluster.  (szetszwo)
 
+    HDFS-8883. NameNode Metrics : Add FSNameSystem lock Queue Length.
+    (Anu Engineer via xyao)
+
     HDFS-8713. Convert DatanodeDescriptor to use SLF4J logging. (wang)
 
   OPTIMIZATIONS

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

@@ -3908,6 +3908,21 @@ public class FSNamesystem implements Namesystem, FSNamesystemMBean,
     return dir.ezManager.getNumEncryptionZones();
   }
 
+  /**
+   * Returns the length of the wait Queue for the FSNameSystemLock.
+   *
+   * A larger number here indicates lots of threads are waiting for
+   * FSNameSystemLock.
+   *
+   * @return int - Number of Threads waiting to acquire FSNameSystemLock
+   */
+  @Override
+  @Metric({"LockQueueLength", "Number of threads waiting to " +
+      "acquire FSNameSystemLock"})
+  public int getFsLockQueueLength() {
+    return fsLock.getQueueLength();
+  }
+
   int getNumberOfDatanodes(DatanodeReportType type) {
     readLock();
     try {

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

@@ -59,4 +59,15 @@ class FSNamesystemLock implements ReadWriteLock {
   public boolean isWriteLockedByCurrentThread() {
     return coarseLock.isWriteLockedByCurrentThread();
   }
+
+  /**
+   * Returns the QueueLength of waiting threads.
+   *
+   * A larger number indicates greater lock contention.
+   *
+   * @return int - Number of threads waiting on this lock
+   */
+  public int getQueueLength() {
+    return coarseLock.getQueueLength();
+  }
 }

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

@@ -189,4 +189,13 @@ public interface FSNamesystemMBean {
    * Return the number of encryption zones in the system.
    */
   int getNumEncryptionZones();
+
+  /**
+   * Returns the length of the wait Queue for the FSNameSystemLock.
+   *
+   * A larger number here indicates lots of threads are waiting for
+   * FSNameSystemLock.
+   * @return int - Number of Threads waiting to acquire FSNameSystemLock
+   */
+  int getFsLockQueueLength();
 }

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

@@ -37,10 +37,15 @@ import org.apache.hadoop.hdfs.server.namenode.ha.HAContext;
 import org.apache.hadoop.hdfs.server.namenode.ha.HAState;
 import org.apache.hadoop.hdfs.server.namenode.snapshot.Snapshot;
 import org.junit.After;
+import org.junit.Assert;
 import org.junit.Test;
 import org.mockito.Mockito;
 import org.mockito.internal.util.reflection.Whitebox;
 
+import java.util.concurrent.CountDownLatch;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+
 public class TestFSNamesystem {
 
   @After
@@ -233,4 +238,27 @@ public class TestFSNamesystem {
     assertEquals(-63,
         FSNamesystem.getEffectiveLayoutVersion(false, -63, -61, -63));
   }
+
+  @Test
+  public void testFSLockGetWaiterCount() throws InterruptedException {
+    final int threadCount = 3;
+    final CountDownLatch latch = new CountDownLatch(threadCount);
+    final FSNamesystemLock rwLock = new FSNamesystemLock(true);
+    rwLock.writeLock().lock();
+    ExecutorService helper = Executors.newFixedThreadPool(threadCount);
+
+    for (int x = 0; x < threadCount; x++) {
+      helper.execute(new Runnable() {
+        @Override
+        public void run() {
+          latch.countDown();
+          rwLock.readLock().lock();
+        }
+      });
+    }
+
+    latch.await();
+    Assert.assertEquals("Expected number of blocked thread not found",
+                        threadCount, rwLock.getQueueLength());
+  }
 }

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

@@ -17,8 +17,8 @@
  */
 package org.apache.hadoop.hdfs.server.namenode;
 
-import org.apache.commons.io.FileUtils;
 import com.google.common.util.concurrent.Uninterruptibles;
+import org.apache.commons.io.FileUtils;
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.FileSystem;
 import org.apache.hadoop.fs.FileUtil;
@@ -376,4 +376,23 @@ public class TestNameNodeMXBean {
       }
     }
   }
+
+  @Test(timeout = 120000)
+  public void testQueueLength() throws Exception {
+    final Configuration conf = new Configuration();
+    MiniDFSCluster cluster = null;
+    try {
+      cluster = new MiniDFSCluster.Builder(conf).numDataNodes(0).build();
+      cluster.waitActive();
+      MBeanServer mbs = ManagementFactory.getPlatformMBeanServer();
+      ObjectName mxbeanNameFs =
+          new ObjectName("Hadoop:service=NameNode,name=FSNamesystem");
+      int queueLength = (int) mbs.getAttribute(mxbeanNameFs, "LockQueueLength");
+      assertEquals(0, queueLength);
+    } finally {
+      if (cluster != null) {
+        cluster.shutdown();
+      }
+    }
+  }
 }