HDFS-9549. TestCacheDirectives#testExceedsCapacity is flaky (Xiao Chen via cmccabe)

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

@@ -2816,6 +2816,9 @@ Release 2.8.0 - UNRELEASED
     HDFS-9842. dfs.datanode.balance.bandwidthPerSec should accept friendly
     size units.  (Lin Yiqun via Arpit Agarwal)
 
+    HDFS-9549. TestCacheDirectives#testExceedsCapacity is flaky (Xiao Chen via
+    cmccabe)
+
 Release 2.7.3 - UNRELEASED
 
   INCOMPATIBLE CHANGES

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

@@ -27,6 +27,7 @@ import java.util.Date;
 import java.util.Iterator;
 import java.util.LinkedList;
 import java.util.List;
+import java.util.Set;
 import java.util.Random;
 import java.util.TreeMap;
 import java.util.concurrent.TimeUnit;
@@ -491,6 +492,26 @@ public class CacheReplicationMonitor extends Thread implements Closeable {
    * Blocks that are over-replicated should be removed from Datanodes.
    */
   private void rescanCachedBlockMap() {
+    // Remove pendingCached blocks that will make DN out-of-capacity.
+    Set<DatanodeDescriptor> datanodes =
+        blockManager.getDatanodeManager().getDatanodes();
+    for (DatanodeDescriptor dn : datanodes) {
+      long remaining = dn.getCacheRemaining();
+      for (Iterator<CachedBlock> it = dn.getPendingCached().iterator();
+           it.hasNext();) {
+        CachedBlock cblock = it.next();
+        BlockInfo blockInfo = blockManager.
+            getStoredBlock(new Block(cblock.getBlockId()));
+        if (blockInfo.getNumBytes() > remaining) {
+          LOG.debug("Block {}: removing from PENDING_CACHED for node {} "
+                  + "because it cannot fit in remaining cache size {}.",
+              cblock.getBlockId(), dn.getDatanodeUuid(), remaining);
+          it.remove();
+        } else {
+          remaining -= blockInfo.getNumBytes();
+        }
+      }
+    }
     for (Iterator<CachedBlock> cbIter = cachedBlocks.iterator();
         cbIter.hasNext(); ) {
       scannedBlocks++;
@@ -531,7 +552,7 @@ public class CacheReplicationMonitor extends Thread implements Closeable {
           DatanodeDescriptor datanode = iter.next();
           datanode.getPendingCached().remove(cblock);
           iter.remove();
-          LOG.trace("Block {}: removing from PENDING_CACHED for node {}"
+          LOG.trace("Block {}: removing from PENDING_CACHED for node {} "
                   + "because we already have {} cached replicas and we only" +
                   " need {}",
               cblock.getBlockId(), datanode.getDatanodeUuid(), numCached,
@@ -686,8 +707,8 @@ public class CacheReplicationMonitor extends Thread implements Closeable {
       long pendingCapacity = pendingBytes + datanode.getCacheRemaining();
       if (pendingCapacity < blockInfo.getNumBytes()) {
         LOG.trace("Block {}: DataNode {} is not a valid possibility " +
-            "because the block has size {}, but the DataNode only has {}" +
-            "bytes of cache remaining ({} pending bytes, {} already cached.",
+            "because the block has size {}, but the DataNode only has {} " +
+            "bytes of cache remaining ({} pending bytes, {} already cached.)",
             blockInfo.getBlockId(), datanode.getDatanodeUuid(),
             blockInfo.getNumBytes(), pendingCapacity, pendingBytes,
             datanode.getCacheRemaining());

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

@@ -413,6 +413,15 @@ public class DatanodeManager {
     return host2DatanodeMap.getDatanodeByXferAddr(host, xferPort);
   }
 
+  /** @return the datanode descriptors for all nodes. */
+  public Set<DatanodeDescriptor> getDatanodes() {
+    final Set<DatanodeDescriptor> datanodes;
+    synchronized (this) {
+      datanodes = new HashSet<>(datanodeMap.values());
+    }
+    return datanodes;
+  }
+
   /** @return the Host2NodesMap */
   public Host2NodesMap getHost2DatanodeMap() {
     return this.host2DatanodeMap;

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

@@ -72,7 +72,6 @@ import org.apache.hadoop.hdfs.protocol.CachePoolStats;
 import org.apache.hadoop.hdfs.protocol.DatanodeInfo;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants.DatanodeReportType;
 import org.apache.hadoop.hdfs.protocol.HdfsConstants.SafeModeAction;
-import org.apache.hadoop.hdfs.server.blockmanagement.CacheReplicationMonitor;
 import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeDescriptor;
 import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeDescriptor.CachedBlocksList.Type;
 import org.apache.hadoop.hdfs.server.blockmanagement.DatanodeManager;
@@ -86,9 +85,6 @@ import org.apache.hadoop.security.AccessControlException;
 import org.apache.hadoop.security.UserGroupInformation;
 import org.apache.hadoop.test.GenericTestUtils;
 import org.apache.hadoop.util.GSet;
-import org.apache.log4j.Level;
-import org.apache.log4j.LogManager;
-import org.apache.log4j.Logger;
 import org.junit.After;
 import org.junit.Assert;
 import org.junit.Before;
@@ -1470,6 +1466,7 @@ public class TestCacheDirectives {
    */
   private void checkPendingCachedEmpty(MiniDFSCluster cluster)
       throws Exception {
+    Thread.sleep(1000);
     cluster.getNamesystem().readLock();
     try {
       final DatanodeManager datanodeManager =
@@ -1501,7 +1498,6 @@ public class TestCacheDirectives {
     waitForCachedBlocks(namenode, -1, numCachedReplicas,
         "testExceeds:1");
     checkPendingCachedEmpty(cluster);
-    Thread.sleep(1000);
     checkPendingCachedEmpty(cluster);
 
     // Try creating a file with giant-sized blocks that exceed cache capacity
@@ -1509,7 +1505,6 @@ public class TestCacheDirectives {
     DFSTestUtil.createFile(dfs, fileName, 4096, fileLen, CACHE_CAPACITY * 2,
         (short) 1, 0xFADED);
     checkPendingCachedEmpty(cluster);
-    Thread.sleep(1000);
     checkPendingCachedEmpty(cluster);
   }