HDFS-3475. Make the replication monitor multipliers configurable. Contributed by Harsh J Chouraria

git-svn-id: https://svn.apache.org/repos/asf/hadoop/common/trunk@1355089 13f79535-47bb-0310-9956-ffa450edef68

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

@@ -250,6 +250,9 @@ Branch-2 ( Unreleased changes )
 
     HDFS-3572. Cleanup code which inits SPNEGO in HttpServer (todd)
 
+    HDFS-3475. Make the replication monitor multipliers configurable.
+    (harsh via eli)
+
   OPTIMIZATIONS
 
     HDFS-2982. Startup performance suffers when there are many edit log

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

@@ -165,6 +165,14 @@ public class DFSConfigKeys extends CommonConfigurationKeys {
   public static final String  DFS_DATANODE_SOCKET_REUSE_KEEPALIVE_KEY = "dfs.datanode.socket.reuse.keepalive";
   public static final int     DFS_DATANODE_SOCKET_REUSE_KEEPALIVE_DEFAULT = 1000;
 
+  // Replication monitoring related keys
+  public static final String DFS_NAMENODE_INVALIDATE_WORK_PCT_PER_ITERATION =
+      "dfs.namenode.invalidate.work.pct.per.iteration";
+  public static final int DFS_NAMENODE_INVALIDATE_WORK_PCT_PER_ITERATION_DEFAULT = 32;
+  public static final String DFS_NAMENODE_REPLICATION_WORK_MULTIPLIER_PER_ITERATION =
+      "dfs.namenode.replication.work.multiplier.per.iteration";
+  public static final int DFS_NAMENODE_REPLICATION_WORK_MULTIPLIER_PER_ITERATION_DEFAULT = 2;
+
   //Delegation token related keys
   public static final String  DFS_NAMENODE_DELEGATION_KEY_UPDATE_INTERVAL_KEY = "dfs.namenode.delegation.key.update-interval";
   public static final long    DFS_NAMENODE_DELEGATION_KEY_UPDATE_INTERVAL_DEFAULT = 24*60*60*1000; // 1 day

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

@@ -68,6 +68,7 @@
 import org.apache.hadoop.util.Daemon;
 
 import com.google.common.annotations.VisibleForTesting;
+import com.google.common.base.Preconditions;
 import com.google.common.collect.Sets;
 
 /**
@@ -193,6 +194,9 @@ public int getPendingDataNodeMessageCount() {
   /** value returned by MAX_CORRUPT_FILES_RETURNED */
   final int maxCorruptFilesReturned;
 
+  final float blocksInvalidateWorkPct;
+  final int blocksReplWorkMultiplier;
+
   /** variable to enable check for enough racks */
   final boolean shouldCheckForEnoughRacks;
 
@@ -245,7 +249,25 @@ public BlockManager(final Namesystem namesystem, final FSClusterStats stats,
     this.maxReplicationStreams = conf.getInt(DFSConfigKeys.DFS_NAMENODE_REPLICATION_MAX_STREAMS_KEY,
                                              DFSConfigKeys.DFS_NAMENODE_REPLICATION_MAX_STREAMS_DEFAULT);
     this.shouldCheckForEnoughRacks = conf.get(DFSConfigKeys.NET_TOPOLOGY_SCRIPT_FILE_NAME_KEY) != null;
-    
+
+    this.blocksInvalidateWorkPct = conf.getFloat(
+        DFSConfigKeys.DFS_NAMENODE_INVALIDATE_WORK_PCT_PER_ITERATION,
+        DFSConfigKeys.DFS_NAMENODE_INVALIDATE_WORK_PCT_PER_ITERATION_DEFAULT);
+    Preconditions.checkArgument(
+        (this.blocksInvalidateWorkPct > 0),
+        DFSConfigKeys.DFS_NAMENODE_INVALIDATE_WORK_PCT_PER_ITERATION +
+        " = '" + this.blocksInvalidateWorkPct + "' is invalid. " +
+        "It should be a positive, non-zero float value " +
+        "indicating a percentage.");
+    this.blocksReplWorkMultiplier = conf.getInt(
+        DFSConfigKeys.DFS_NAMENODE_REPLICATION_WORK_MULTIPLIER_PER_ITERATION,
+        DFSConfigKeys.DFS_NAMENODE_REPLICATION_WORK_MULTIPLIER_PER_ITERATION_DEFAULT);
+    Preconditions.checkArgument(
+        (this.blocksReplWorkMultiplier > 0),
+        DFSConfigKeys.DFS_NAMENODE_REPLICATION_WORK_MULTIPLIER_PER_ITERATION +
+        " = '" + this.blocksReplWorkMultiplier + "' is invalid. " +
+        "It should be a positive, non-zero integer value.");
+
     this.replicationRecheckInterval = 
       conf.getInt(DFSConfigKeys.DFS_NAMENODE_REPLICATION_INTERVAL_KEY, 
                   DFSConfigKeys.DFS_NAMENODE_REPLICATION_INTERVAL_DEFAULT) * 1000L;
@@ -2897,8 +2919,6 @@ public int numOfUnderReplicatedBlocks() {
    * Periodically calls computeReplicationWork().
    */
   private class ReplicationMonitor implements Runnable {
-    private static final int INVALIDATE_WORK_PCT_PER_ITERATION = 32;
-    private static final int REPLICATION_WORK_MULTIPLIER_PER_ITERATION = 2;
 
     @Override
     public void run() {
@@ -2938,9 +2958,9 @@ int computeDatanodeWork() {
 
     final int numlive = heartbeatManager.getLiveDatanodeCount();
     final int blocksToProcess = numlive
-        * ReplicationMonitor.REPLICATION_WORK_MULTIPLIER_PER_ITERATION;
+        * this.blocksReplWorkMultiplier;
     final int nodesToProcess = (int) Math.ceil(numlive
-        * ReplicationMonitor.INVALIDATE_WORK_PCT_PER_ITERATION / 100.0);
+        * this.blocksInvalidateWorkPct);
 
     int workFound = this.computeReplicationWork(blocksToProcess);
 

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

@@ -870,5 +870,35 @@
   <value>${dfs.web.authentication.kerberos.principal}</value>
 </property>
 
+<property>
+  <name>dfs.namenode.invalidate.work.pct.per.iteration</name>
+  <value>0.32f</value>
+  <description>
+    *Note*: Advanced property. Change with caution.
+    This determines the percentage amount of block
+    invalidations (deletes) to do over a single DN heartbeat
+    deletion command. The final deletion count is determined by applying this
+    percentage to the number of live nodes in the system.
+    The resultant number is the number of blocks from the deletion list
+    chosen for proper invalidation over a single heartbeat of a single DN.
+    Value should be a positive, non-zero percentage in float notation (X.Yf),
+    with 1.0f meaning 100%.
+  </description>
+</property>
+
+<property>
+  <name>dfs.namenode.replication.work.multiplier.per.iteration</name>
+  <value>2</value>
+  <description>
+    *Note*: Advanced property. Change with caution.
+    This determines the total amount of block transfers to begin in
+    parallel at a DN, for replication, when such a command list is being
+    sent over a DN heartbeat by the NN. The actual number is obtained by
+    multiplying this multiplier with the total number of live nodes in the
+    cluster. The result number is the number of blocks to begin transfers
+    immediately for, per DN heartbeat. This number can be any positive,
+    non-zero integer.
+  </description>
+</property>
 
 </configuration>