hbase-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From jxi...@apache.org
Subject git commit: HBASE-12007 StochasticBalancer should avoid putting user regions on master
Date Fri, 19 Sep 2014 18:42:55 GMT
Repository: hbase
Updated Branches:
  refs/heads/branch-1 73cd983e7 -> 57f6ff508


HBASE-12007 StochasticBalancer should avoid putting user regions on master


Project: http://git-wip-us.apache.org/repos/asf/hbase/repo
Commit: http://git-wip-us.apache.org/repos/asf/hbase/commit/57f6ff50
Tree: http://git-wip-us.apache.org/repos/asf/hbase/tree/57f6ff50
Diff: http://git-wip-us.apache.org/repos/asf/hbase/diff/57f6ff50

Branch: refs/heads/branch-1
Commit: 57f6ff50893583ed7c8938c50eb67b363fbe7f30
Parents: 73cd983
Author: Jimmy Xiang <jxiang@cloudera.com>
Authored: Wed Sep 17 12:27:48 2014 -0700
Committer: Jimmy Xiang <jxiang@cloudera.com>
Committed: Fri Sep 19 11:42:00 2014 -0700

----------------------------------------------------------------------
 .../hbase/master/balancer/BaseLoadBalancer.java | 17 ++------
 .../master/balancer/StochasticLoadBalancer.java | 42 +++++++++++++-------
 .../hadoop/hbase/TestRegionRebalancing.java     | 19 +++------
 .../balancer/TestStochasticLoadBalancer.java    |  3 +-
 4 files changed, 39 insertions(+), 42 deletions(-)
----------------------------------------------------------------------


http://git-wip-us.apache.org/repos/asf/hbase/blob/57f6ff50/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/BaseLoadBalancer.java
----------------------------------------------------------------------
diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/BaseLoadBalancer.java
b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/BaseLoadBalancer.java
index 8af5c88..79bc91b 100644
--- a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/BaseLoadBalancer.java
+++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/BaseLoadBalancer.java
@@ -158,6 +158,7 @@ public abstract class BaseLoadBalancer implements LoadBalancer {
         tablesOnMaster, rackManager);
     }
 
+    @SuppressWarnings("unchecked")
     protected Cluster(
         ServerName masterServerName,
         Collection<HRegionInfo> unassignedRegions,
@@ -845,29 +846,19 @@ public abstract class BaseLoadBalancer implements LoadBalancer {
   private static final Random RANDOM = new Random(System.currentTimeMillis());
   private static final Log LOG = LogFactory.getLog(BaseLoadBalancer.class);
 
-  // The weight means that each region on the active/backup master is
+  // The weight means that each region on the backup master is
   // equal to that many regions on a normal regionserver, in calculating
-  // the region load by the load balancer. So that the active/backup master
+  // the region load by the load balancer. So that the backup master
   // can host less (or equal if weight = 1) regions than normal regionservers.
   //
   // The weight can be used to control the number of regions on backup
   // masters, which shouldn't host as many regions as normal regionservers.
   // So that we don't need to move around too many regions when a
   // backup master becomes the active one.
-  //
-  // Currently, the active master weight is used only by StockasticLoadBalancer.
-  // Generally, we don't put any user regions on the active master, which
-  // only hosts regions of tables defined in TABLES_ON_MASTER.
-  // That's why the default activeMasterWeight is high.
   public static final String BACKUP_MASTER_WEIGHT_KEY =
     "hbase.balancer.backupMasterWeight";
   public static final int DEFAULT_BACKUP_MASTER_WEIGHT = 0;
 
-  private static final String ACTIVE_MASTER_WEIGHT_KEY =
-    "hbase.balancer.activeMasterWeight";
-  private static final int DEFAULT_ACTIVE_MASTER_WEIGHT = 200;
-
-  protected int activeMasterWeight;
   protected int backupMasterWeight;
 
   // a flag to indicate if assigning regions to backup masters
@@ -888,8 +879,6 @@ public abstract class BaseLoadBalancer implements LoadBalancer {
     else if (slop > 1) slop = 1;
 
     this.config = conf;
-    activeMasterWeight = conf.getInt(
-      ACTIVE_MASTER_WEIGHT_KEY, DEFAULT_ACTIVE_MASTER_WEIGHT);
     backupMasterWeight = conf.getInt(
       BACKUP_MASTER_WEIGHT_KEY, DEFAULT_BACKUP_MASTER_WEIGHT);
     if (backupMasterWeight < 1) {

http://git-wip-us.apache.org/repos/asf/hbase/blob/57f6ff50/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/StochasticLoadBalancer.java
----------------------------------------------------------------------
diff --git a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/StochasticLoadBalancer.java
b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/StochasticLoadBalancer.java
index d32094d..d597843 100644
--- a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/StochasticLoadBalancer.java
+++ b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/StochasticLoadBalancer.java
@@ -157,7 +157,7 @@ public class StochasticLoadBalancer extends BaseLoadBalancer {
     regionReplicaRackCostFunction = new RegionReplicaRackCostFunction(conf);
 
     costFunctions = new CostFunction[]{
-      new RegionCountSkewCostFunction(conf, activeMasterWeight, backupMasterWeight),
+      new RegionCountSkewCostFunction(conf, backupMasterWeight),
       new MoveCostFunction(conf),
       localityCost,
       new TableSkewCostFunction(conf),
@@ -422,7 +422,11 @@ public class StochasticLoadBalancer extends BaseLoadBalancer {
         return -1;
       }
 
-      return RANDOM.nextInt(cluster.numServers);
+      int n = RANDOM.nextInt(cluster.numServers);
+      if (cluster.numServers > 1 && cluster.isActiveMaster(n)) {
+        n = (n + 1) % cluster.numServers;
+      }
+      return n;
     }
 
     protected int pickRandomRack(Cluster cluster) {
@@ -434,7 +438,7 @@ public class StochasticLoadBalancer extends BaseLoadBalancer {
     }
 
     protected int pickOtherRandomServer(Cluster cluster, int serverIndex) {
-      if (cluster.numServers < 2) {
+      if (cluster.numServers <= 2) {
         return -1;
       }
       while (true) {
@@ -524,8 +528,13 @@ public class StochasticLoadBalancer extends BaseLoadBalancer {
     private int pickLeastLoadedServer(final Cluster cluster, int thisServer) {
       Integer[] servers = cluster.serverIndicesSortedByRegionCount;
 
+      if (servers.length <= 2) {
+        return -1;
+      }
+
       int index = 0;
-      while (servers[index] == null || servers[index] == thisServer) {
+      while (servers[index] == null || servers[index] == thisServer
+          || cluster.isActiveMaster(index)) {
         index++;
         if (index == servers.length) {
           return -1;
@@ -538,7 +547,8 @@ public class StochasticLoadBalancer extends BaseLoadBalancer {
       Integer[] servers = cluster.serverIndicesSortedByRegionCount;
 
       int index = servers.length - 1;
-      while (servers[index] == null || servers[index] == thisServer) {
+      while (servers[index] == null || servers[index] == thisServer
+          || cluster.isActiveMaster(index)) {
         index--;
         if (index < 0) {
           return -1;
@@ -788,14 +798,23 @@ public class StochasticLoadBalancer extends BaseLoadBalancer {
     protected double costFromArray(double[] stats) {
       double totalCost = 0;
       double total = getSum(stats);
-      double mean = total/((double)stats.length);
+
       double count = stats.length;
+      if (stats.length > 1 && cluster.masterServerName != null) {
+        count--; // Exclude the active master
+      }
+      double mean = total/count;
 
       // Compute max as if all region servers had 0 and one had the sum of all costs.  This
must be
       // a zero sum cost for this to make sense.
       // TODO: Should we make this sum of square errors?
       double max = ((count - 1) * mean) + (total - mean);
-      for (double n : stats) {
+      for (int i=0; i<stats.length; i++) {
+        if (stats.length > 1 && cluster.isActiveMaster(i)) {
+          // Not count the active master load
+          continue;
+        }
+        double n = stats[i];
         double diff = Math.abs(mean - n);
         totalCost += diff;
       }
@@ -882,16 +901,13 @@ public class StochasticLoadBalancer extends BaseLoadBalancer {
         "hbase.master.balancer.stochastic.regionCountCost";
     private static final float DEFAULT_REGION_COUNT_SKEW_COST = 500;
 
-    private double activeMasterWeight;
     private double backupMasterWeight;
     private double[] stats = null;
 
-    RegionCountSkewCostFunction(Configuration conf,
-        double activeMasterWeight, double backupMasterWeight) {
+    RegionCountSkewCostFunction(Configuration conf, double backupMasterWeight) {
       super(conf);
       // Load multiplier should be the greatest as it is the most general way to balance
data.
       this.setMultiplier(conf.getFloat(REGION_COUNT_SKEW_COST_KEY, DEFAULT_REGION_COUNT_SKEW_COST));
-      this.activeMasterWeight = activeMasterWeight;
       this.backupMasterWeight = backupMasterWeight;
     }
 
@@ -905,9 +921,7 @@ public class StochasticLoadBalancer extends BaseLoadBalancer {
         stats[i] = cluster.regionsPerServer[i].length;
         // Use some weight on regions assigned to active/backup masters,
         // so that they won't carry as many regions as normal regionservers.
-        if (cluster.isActiveMaster(i)) {
-          stats[i] += cluster.numUserRegionsOnMaster * (activeMasterWeight - 1);
-        } else if (cluster.isBackupMaster(i)) {
+        if (cluster.isBackupMaster(i)) {
           stats[i] *= backupMasterWeight;
         }
       }

http://git-wip-us.apache.org/repos/asf/hbase/blob/57f6ff50/hbase-server/src/test/java/org/apache/hadoop/hbase/TestRegionRebalancing.java
----------------------------------------------------------------------
diff --git a/hbase-server/src/test/java/org/apache/hadoop/hbase/TestRegionRebalancing.java
b/hbase-server/src/test/java/org/apache/hadoop/hbase/TestRegionRebalancing.java
index b3fa6c1..8509be9 100644
--- a/hbase-server/src/test/java/org/apache/hadoop/hbase/TestRegionRebalancing.java
+++ b/hbase-server/src/test/java/org/apache/hadoop/hbase/TestRegionRebalancing.java
@@ -29,7 +29,6 @@ import java.util.List;
 
 import org.apache.commons.logging.Log;
 import org.apache.commons.logging.LogFactory;
-import org.apache.hadoop.hbase.client.Admin;
 import org.apache.hadoop.hbase.client.HBaseAdmin;
 import org.apache.hadoop.hbase.client.HTable;
 import org.apache.hadoop.hbase.client.RegionLocator;
@@ -94,6 +93,7 @@ public class TestRegionRebalancing {
    * @throws InterruptedException
    */
   @Test (timeout=300000)
+  @SuppressWarnings("deprecation")
   public void testRebalanceOnRegionServerNumberChange()
   throws IOException, InterruptedException {
     HBaseAdmin admin = new HBaseAdmin(UTIL.getConfiguration());
@@ -147,15 +147,7 @@ public class TestRegionRebalancing {
     assert(UTIL.getHBaseCluster().getMaster().balance() == true);
     assertRegionsAreBalanced();
     table.close();
-  }
-
-  /** figure out how many regions are currently being served. */
-  private int getRegionCount() throws IOException {
-    int total = 0; // Regions on master are ignored since not counted for balancing
-    for (HRegionServer server : getOnlineRegionServers()) {
-      total += ProtobufUtil.getOnlineRegions(server.getRSRpcServices()).size();
-    }
-    return total;
+    admin.close();
   }
 
   /**
@@ -175,7 +167,7 @@ public class TestRegionRebalancing {
       // make sure all the regions are reassigned before we test balance
       waitForAllRegionsAssigned();
 
-      int regionCount = getRegionCount();
+      long regionCount = UTIL.getMiniHBaseCluster().countServedRegions();
       List<HRegionServer> servers = getOnlineRegionServers();
       double avg = UTIL.getHBaseCluster().getMaster().getAverageLoad();
       int avgLoadPlusSlop = (int)Math.ceil(avg * (1 + slop));
@@ -239,9 +231,10 @@ public class TestRegionRebalancing {
    */
   private void waitForAllRegionsAssigned() throws IOException {
     int totalRegions = HBaseTestingUtility.KEYS.length;
-    while (getRegionCount() < totalRegions) {
+    while (UTIL.getMiniHBaseCluster().countServedRegions() < totalRegions) {
     // while (!cluster.getMaster().allRegionsAssigned()) {
-      LOG.debug("Waiting for there to be "+ totalRegions +" regions, but there are " + getRegionCount()
+ " right now.");
+      LOG.debug("Waiting for there to be "+ totalRegions +" regions, but there are "
+        + UTIL.getMiniHBaseCluster().countServedRegions() + " right now.");
       try {
         Thread.sleep(200);
       } catch (InterruptedException e) {}

http://git-wip-us.apache.org/repos/asf/hbase/blob/57f6ff50/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancer.java
----------------------------------------------------------------------
diff --git a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancer.java
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancer.java
index bc33cce..7c7ac49 100644
--- a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancer.java
+++ b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancer.java
@@ -193,7 +193,7 @@ public class TestStochasticLoadBalancer extends BalancerTestBase {
   public void testSkewCost() {
     Configuration conf = HBaseConfiguration.create();
     StochasticLoadBalancer.CostFunction
-        costFunction = new StochasticLoadBalancer.RegionCountSkewCostFunction(conf, 1, 1);
+        costFunction = new StochasticLoadBalancer.RegionCountSkewCostFunction(conf, 1);
     for (int[] mockCluster : clusterStateMocks) {
       costFunction.init(mockCluster(mockCluster));
       double cost = costFunction.cost();
@@ -239,6 +239,7 @@ public class TestStochasticLoadBalancer extends BalancerTestBase {
     Configuration conf = HBaseConfiguration.create();
     StochasticLoadBalancer.CostFromRegionLoadFunction
         costFunction = new StochasticLoadBalancer.MemstoreSizeCostFunction(conf);
+    costFunction.init(mockCluster(new int[]{0, 0, 0, 0, 1}));
 
     double[] statOne = new double[100];
     for (int i =0; i < 100; i++) {


Mime
View raw message