HDFS-9107. Prevent NN's unrecoverable death spiral after full GC (Daryn Sharp via Colin P. McCabe)

(cherry-picked from 4e7c6a653f108d44589f84d78a03d92ee0e8a3c3)

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

@@ -1071,6 +1071,9 @@ Release 2.8.0 - UNRELEASED
     HDFS-9123. Copying from the root to a subdirectory should be forbidden.
     (Wei-Chiu Chuang via Yongjun Zhang)
 
+    HDFS-9107. Prevent NN's unrecoverable death spiral after full GC (Daryn
+    Sharp via Colin P. McCabe)
+
 Release 2.7.2 - UNRELEASED
 
   INCOMPATIBLE CHANGES

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

@@ -24,6 +24,7 @@ import java.util.IdentityHashMap;
 import java.util.List;
 import java.util.Map;
 import java.util.Set;
+import java.util.concurrent.TimeUnit;
 
 import org.apache.hadoop.conf.Configuration;
 import org.apache.hadoop.fs.StorageType;
@@ -34,10 +35,13 @@ import org.apache.hadoop.hdfs.server.namenode.Namesystem;
 import org.apache.hadoop.hdfs.server.protocol.StorageReport;
 import org.apache.hadoop.hdfs.server.protocol.VolumeFailureSummary;
 import org.apache.hadoop.util.Daemon;
+import org.apache.hadoop.util.StopWatch;
 import org.apache.hadoop.util.Time;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
+import com.google.common.annotations.VisibleForTesting;
+
 /**
  * Manage the heartbeats received from datanodes.
  * The datanode list and statistics are synchronized
@@ -62,8 +66,8 @@ class HeartbeatManager implements DatanodeStatistics {
   private final long heartbeatRecheckInterval;
   /** Heartbeat monitor thread */
   private final Daemon heartbeatThread = new Daemon(new Monitor());
+  private final StopWatch heartbeatStopWatch = new StopWatch();
 
-    
   final Namesystem namesystem;
   final BlockManager blockManager;
 
@@ -260,7 +264,18 @@ class HeartbeatManager implements DatanodeStatistics {
       stats.add(node);
     }
   }
-  
+
+  @VisibleForTesting
+  void restartHeartbeatStopWatch() {
+    heartbeatStopWatch.reset().start();
+  }
+
+  @VisibleForTesting
+  boolean shouldAbortHeartbeatCheck(long offset) {
+    long elapsed = heartbeatStopWatch.now(TimeUnit.MILLISECONDS);
+    return elapsed + offset > heartbeatRecheckInterval;
+  }
+
   /**
    * Check if there are any expired heartbeats, and if so,
    * whether any blocks have to be re-replicated.
@@ -307,6 +322,10 @@ class HeartbeatManager implements DatanodeStatistics {
       int numOfStaleStorages = 0;
       synchronized(this) {
         for (DatanodeDescriptor d : datanodes) {
+          // check if an excessive GC pause has occurred
+          if (shouldAbortHeartbeatCheck(0)) {
+            return;
+          }
           if (dead == null && dm.isDatanodeDead(d)) {
             stats.incrExpiredHeartbeats();
             dead = d;
@@ -375,6 +394,7 @@ class HeartbeatManager implements DatanodeStatistics {
     @Override
     public void run() {
       while(namesystem.isRunning()) {
+        restartHeartbeatStopWatch();
         try {
           final long now = Time.monotonicNow();
           if (lastHeartbeatCheck + heartbeatRecheckInterval < now) {
@@ -396,6 +416,12 @@ class HeartbeatManager implements DatanodeStatistics {
           Thread.sleep(5000);  // 5 seconds
         } catch (InterruptedException ie) {
         }
+        // avoid declaring nodes dead for another cycle if a GC pause lasts
+        // longer than the node recheck interval
+        if (shouldAbortHeartbeatCheck(-5000)) {
+          LOG.warn("Skipping next heartbeat scan due to excessive pause");
+          lastHeartbeatCheck = Time.monotonicNow();
+        }
       }
     }
   }

hadoop-hdfs-project/hadoop-hdfs/src/test/java/org/apache/hadoop/hdfs/server/blockmanagement/TestHeartbeatHandling.java

@@ -18,6 +18,8 @@
 package org.apache.hadoop.hdfs.server.blockmanagement;
 
 import static org.junit.Assert.assertEquals;
+import static org.junit.Assert.assertFalse;
+import static org.junit.Assert.assertTrue;
 
 import java.util.ArrayList;
 
@@ -33,6 +35,7 @@ import org.apache.hadoop.hdfs.server.common.HdfsServerConstants.BlockUCState;
 import org.apache.hadoop.hdfs.server.datanode.DataNodeTestUtils;
 import org.apache.hadoop.hdfs.server.namenode.FSNamesystem;
 import org.apache.hadoop.hdfs.server.namenode.NameNodeAdapter;
+import org.apache.hadoop.hdfs.server.namenode.Namesystem;
 import org.apache.hadoop.hdfs.server.protocol.BlockCommand;
 import org.apache.hadoop.hdfs.server.protocol.BlockRecoveryCommand;
 import org.apache.hadoop.hdfs.server.protocol.DatanodeCommand;
@@ -40,6 +43,7 @@ import org.apache.hadoop.hdfs.server.protocol.DatanodeProtocol;
 import org.apache.hadoop.hdfs.server.protocol.DatanodeRegistration;
 import org.apache.hadoop.hdfs.server.protocol.DatanodeStorage;
 import org.junit.Test;
+import org.mockito.Mockito;
 
 /**
  * Test if FSNamesystem handles heartbeat right
@@ -243,4 +247,27 @@ public class TestHeartbeatHandling {
       cluster.shutdown();
     }
   }
+
+  @Test
+  public void testHeartbeatStopWatch() throws Exception {
+   Namesystem ns = Mockito.mock(Namesystem.class);
+   BlockManager bm = Mockito.mock(BlockManager.class);
+   Configuration conf = new Configuration();
+   long recheck = 2000;
+   conf.setLong(
+       DFSConfigKeys.DFS_NAMENODE_HEARTBEAT_RECHECK_INTERVAL_KEY, recheck);
+   HeartbeatManager monitor = new HeartbeatManager(ns, bm, conf);
+   monitor.restartHeartbeatStopWatch();
+   assertFalse(monitor.shouldAbortHeartbeatCheck(0));
+   // sleep shorter than recheck and verify shouldn't abort
+   Thread.sleep(100);
+   assertFalse(monitor.shouldAbortHeartbeatCheck(0));
+   // sleep longer than recheck and verify should abort unless ignore delay
+   Thread.sleep(recheck);
+   assertTrue(monitor.shouldAbortHeartbeatCheck(0));
+   assertFalse(monitor.shouldAbortHeartbeatCheck(-recheck*3));
+   // ensure it resets properly
+   monitor.restartHeartbeatStopWatch();
+   assertFalse(monitor.shouldAbortHeartbeatCheck(0));
+  }
 }