This is an automated email from the ASF dual-hosted git repository.
stack pushed a commit to branch branch-2
in repository https://gitbox.apache.org/repos/asf/hbase.git
The following commit(s) were added to refs/heads/branch-2 by this push:
new dfc9ac8 HBASE-25739 TableSkewCostFunction need to use aggregated
deviation - backport
dfc9ac8 is described below
commit dfc9ac8ffb83b914af27f0e6096bc46d9d87b95f
Author: Clara Xiong <clarax98...@gmail.com>
AuthorDate: Mon Jul 12 15:45:11 2021 -0700
HBASE-25739 TableSkewCostFunction need to use aggregated deviation -
backport
Signed-off-by: Duo Zhang <zhang...@apache.org>
Signed-off-by: stack <st...@duboce.net>
Reviewed-by: Nick Dimiduk <ndimi...@apache.org>
---
.../master/balancer/BalancerClusterState.java | 57 ++++++++++++++--------
.../hadoop/hbase/master/balancer/CostFunction.java | 9 ++--
.../hbase/master/balancer/DoubleArrayCost.java | 51 +++++++++++--------
.../master/balancer/StochasticLoadBalancer.java | 5 +-
.../master/balancer/TableSkewCostFunction.java | 15 +++---
.../hbase/master/balancer/BalancerTestBase.java | 2 +-
.../master/balancer/TestBaseLoadBalancer.java | 4 +-
.../TestStochasticLoadBalancerBalanceCluster.java | 3 +-
...estStochasticLoadBalancerHeterogeneousCost.java | 1 -
.../TestStochasticLoadBalancerLargeCluster.java | 3 ++
10 files changed, 91 insertions(+), 59 deletions(-)
diff --git
a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/BalancerClusterState.java
b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/BalancerClusterState.java
index 7fa92cb..0032002 100644
---
a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/BalancerClusterState.java
+++
b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/BalancerClusterState.java
@@ -81,6 +81,11 @@ class BalancerClusterState {
int[] initialRegionIndexToServerIndex; // regionIndex -> serverIndex
(initial cluster state)
int[] regionIndexToTableIndex; // regionIndex -> tableIndex
int[][] numRegionsPerServerPerTable; // serverIndex -> tableIndex -> #
regions
+ int[] numRegionsPerTable; // tableIndex -> region count
+ double[] meanRegionsPerTable; // mean region count per table
+ double[] regionSkewByTable; // skew on RS per by table
+ double[] minRegionSkewByTable; // min skew on RS per by table
+ double[] maxRegionSkewByTable; // max skew on RS per by table
int[] numMaxRegionsPerTable; // tableIndex -> max number of regions in a
single RS
int[] regionIndexToPrimaryIndex; // regionIndex -> regionIndex of the primary
boolean hasRegionReplicas = false; // whether there is regions with replicas
@@ -290,7 +295,9 @@ class BalancerClusterState {
}
numTables = tables.size();
+ LOG.debug("Number of tables={}", numTables);
numRegionsPerServerPerTable = new int[numServers][numTables];
+ numRegionsPerTable = new int[numTables];
for (int i = 0; i < numServers; i++) {
for (int j = 0; j < numTables; j++) {
@@ -301,15 +308,26 @@ class BalancerClusterState {
for (int i = 0; i < regionIndexToServerIndex.length; i++) {
if (regionIndexToServerIndex[i] >= 0) {
numRegionsPerServerPerTable[regionIndexToServerIndex[i]][regionIndexToTableIndex[i]]++;
+ numRegionsPerTable[regionIndexToTableIndex[i]]++;
}
}
- numMaxRegionsPerTable = new int[numTables];
+ // Avoid repeated computation for planning
+ meanRegionsPerTable = new double[numTables];
+ regionSkewByTable = new double[numTables];
+ maxRegionSkewByTable = new double[numTables];
+ minRegionSkewByTable = new double[numTables];
+
+ for (int i = 0; i < numTables; i++) {
+ meanRegionsPerTable[i] = Double.valueOf(numRegionsPerTable[i]) /
numServers;
+ minRegionSkewByTable[i] +=
DoubleArrayCost.getMinSkew(numRegionsPerTable[i], numServers);
+ maxRegionSkewByTable[i] +=
DoubleArrayCost.getMaxSkew(numRegionsPerTable[i], numServers);
+ }
+
for (int[] aNumRegionsPerServerPerTable : numRegionsPerServerPerTable) {
- for (tableIndex = 0; tableIndex < aNumRegionsPerServerPerTable.length;
tableIndex++) {
- if (aNumRegionsPerServerPerTable[tableIndex] >
numMaxRegionsPerTable[tableIndex]) {
- numMaxRegionsPerTable[tableIndex] =
aNumRegionsPerServerPerTable[tableIndex];
- }
+ for (int tableIdx = 0; tableIdx < aNumRegionsPerServerPerTable.length;
tableIdx++) {
+ regionSkewByTable[tableIdx] +=
+ Math.abs(aNumRegionsPerServerPerTable[tableIdx] -
meanRegionsPerTable[tableIdx]);
}
}
@@ -671,22 +689,13 @@ class BalancerClusterState {
int tableIndex = regionIndexToTableIndex[region];
if (oldServer >= 0) {
numRegionsPerServerPerTable[oldServer][tableIndex]--;
+ // update regionSkewPerTable for the move from old server
+ regionSkewByTable[tableIndex] += getSkewChangeFor(oldServer, tableIndex,
-1);
}
numRegionsPerServerPerTable[newServer][tableIndex]++;
- // check whether this caused maxRegionsPerTable in the new Server to be
updated
- if (numRegionsPerServerPerTable[newServer][tableIndex] >
numMaxRegionsPerTable[tableIndex]) {
- numMaxRegionsPerTable[tableIndex] =
numRegionsPerServerPerTable[newServer][tableIndex];
- } else if (oldServer >= 0 &&
(numRegionsPerServerPerTable[oldServer][tableIndex]
- + 1) == numMaxRegionsPerTable[tableIndex]) {
- // recompute maxRegionsPerTable since the previous value was coming from
the old server
- numMaxRegionsPerTable[tableIndex] = 0;
- for (int[] aNumRegionsPerServerPerTable : numRegionsPerServerPerTable) {
- if (aNumRegionsPerServerPerTable[tableIndex] >
numMaxRegionsPerTable[tableIndex]) {
- numMaxRegionsPerTable[tableIndex] =
aNumRegionsPerServerPerTable[tableIndex];
- }
- }
- }
+ // update regionSkewPerTable for the move to new server
+ regionSkewByTable[tableIndex] += getSkewChangeFor(newServer, tableIndex,
1);
// update for servers
int primary = regionIndexToPrimaryIndex[region];
@@ -856,10 +865,18 @@ class BalancerClusterState {
.append(Arrays.toString(serverIndicesSortedByRegionCount)).append(",
regionsPerServer=")
.append(Arrays.deepToString(regionsPerServer));
- desc.append(",
numMaxRegionsPerTable=").append(Arrays.toString(numMaxRegionsPerTable))
+ desc.append(",
regionSkewByTable=").append(Arrays.toString(regionSkewByTable))
.append(", numRegions=").append(numRegions).append(",
numServers=").append(numServers)
.append(", numTables=").append(numTables).append(",
numMovedRegions=").append(numMovedRegions)
.append('}');
return desc.toString();
}
-}
\ No newline at end of file
+
+ private double getSkewChangeFor(int serverIndex, int tableIndex, double
regionCountChange) {
+ double curSkew =
Math.abs(numRegionsPerServerPerTable[serverIndex][tableIndex] -
+ meanRegionsPerTable[tableIndex]);
+ double oldSkew =
Math.abs(numRegionsPerServerPerTable[serverIndex][tableIndex] -
+ regionCountChange - meanRegionsPerTable[tableIndex]);
+ return curSkew - oldSkew;
+ }
+}
diff --git
a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/CostFunction.java
b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/CostFunction.java
index ecbf400..2735b69 100644
---
a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/CostFunction.java
+++
b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/CostFunction.java
@@ -25,6 +25,8 @@ import org.apache.yetus.audience.InterfaceAudience;
@InterfaceAudience.Private
abstract class CostFunction {
+ public static final double COST_EPSILON = 0.0001;
+
private float multiplier = 0;
protected BalancerClusterState cluster;
@@ -89,13 +91,14 @@ abstract class CostFunction {
* @return The scaled value.
*/
protected static double scale(double min, double max, double value) {
- if (max <= min || value <= min) {
+ if (max <= min || value <= min
+ || Math.abs(max - min) <= COST_EPSILON || Math.abs(value - min) <=
COST_EPSILON) {
return 0;
}
- if ((max - min) == 0) {
+ if (max <= min || Math.abs(max - min) <= COST_EPSILON) {
return 0;
}
return Math.max(0d, Math.min(1d, (value - min) / (max - min)));
}
-}
\ No newline at end of file
+}
diff --git
a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/DoubleArrayCost.java
b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/DoubleArrayCost.java
index 11325c0..35dcdeb 100644
---
a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/DoubleArrayCost.java
+++
b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/DoubleArrayCost.java
@@ -72,31 +72,14 @@ final class DoubleArrayCost {
double count = stats.length;
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.
- double max = ((count - 1) * mean) + (total - mean);
-
- // It's possible that there aren't enough regions to go around
- double min;
- if (count > total) {
- min = ((count - total) * mean) + ((1 - mean) * total);
- } else {
- // Some will have 1 more than everything else.
- int numHigh = (int) (total - (Math.floor(mean) * count));
- int numLow = (int) (count - numHigh);
-
- min = (numHigh * (Math.ceil(mean) - mean)) + (numLow * (mean -
Math.floor(mean)));
-
- }
- min = Math.max(0, min);
for (int i = 0; i < stats.length; i++) {
double n = stats[i];
double diff = Math.abs(mean - n);
totalCost += diff;
}
- double scaled = CostFunction.scale(min, max, totalCost);
- return scaled;
+ return CostFunction.scale(getMinSkew(total, count),
+ getMaxSkew(total, count), totalCost);
}
private static double getSum(double[] stats) {
@@ -106,4 +89,34 @@ final class DoubleArrayCost {
}
return total;
}
+
+ /**
+ * Return the min skew of distribution
+ * @param total is total number of regions
+ */
+ public static double getMinSkew(double total, double numServers) {
+ double mean = total / numServers;
+ // It's possible that there aren't enough regions to go around
+ double min;
+ if (numServers > total) {
+ min = ((numServers - total) * mean + (1 - mean) * total) ;
+ } else {
+ // Some will have 1 more than everything else.
+ int numHigh = (int) (total - (Math.floor(mean) * numServers));
+ int numLow = (int) (numServers - numHigh);
+ min = numHigh * (Math.ceil(mean) - mean) + numLow * (mean -
Math.floor(mean));
+ }
+ return min;
+ }
+
+ /**
+ * Return the max deviation of distribution
+ * 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.
+ * @param total is total number of regions
+ */
+ public static double getMaxSkew(double total, double numServers) {
+ double mean = total / numServers;
+ return (total - mean) + (numServers - 1) * mean;
+ }
}
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 a24c78e..ba214b7 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
@@ -127,7 +127,7 @@ public class StochasticLoadBalancer extends
BaseLoadBalancer {
private int stepsPerRegion = 800;
private long maxRunningTime = 30 * 1000 * 1; // 30 seconds.
private int numRegionLoadsToRemember = 15;
- private float minCostNeedBalance = 0.05f;
+ private float minCostNeedBalance = 0.025f;
private boolean isBalancerDecisionRecording = false;
private boolean isBalancerRejectionRecording = false;
@@ -259,7 +259,8 @@ public class StochasticLoadBalancer extends
BaseLoadBalancer {
this.namedQueueRecorder = NamedQueueRecorder.getInstance(conf);
}
- LOG.info("Loaded config; maxSteps=" + maxSteps + ", stepsPerRegion=" +
stepsPerRegion +
+ LOG.info("Loaded config; maxSteps=" + maxSteps + ", runMaxSteps=" +
runMaxSteps +
+ ", stepsPerRegion=" + stepsPerRegion +
", maxRunningTime=" + maxRunningTime + ", isByTable=" + isByTable + ",
CostFunctions=" +
Arrays.toString(getCostFunctionNames()) + " etc.");
}
diff --git
a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/TableSkewCostFunction.java
b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/TableSkewCostFunction.java
index 4ab0228..efd7983 100644
---
a/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/TableSkewCostFunction.java
+++
b/hbase-server/src/main/java/org/apache/hadoop/hbase/master/balancer/TableSkewCostFunction.java
@@ -37,14 +37,11 @@ class TableSkewCostFunction extends CostFunction {
@Override
protected double cost() {
- double max = cluster.numRegions;
- double min = ((double) cluster.numRegions) / cluster.numServers;
- double value = 0;
-
- for (int i = 0; i < cluster.numMaxRegionsPerTable.length; i++) {
- value += cluster.numMaxRegionsPerTable[i];
+ double cost = 0;
+ for (int tableIdx = 0; tableIdx < cluster.numTables; tableIdx++) {
+ cost += scale(cluster.minRegionSkewByTable[tableIdx],
+ cluster.maxRegionSkewByTable[tableIdx],
cluster.regionSkewByTable[tableIdx]);
}
-
- return scale(min, max, value);
+ return cost;
}
-}
\ No newline at end of file
+}
diff --git
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/BalancerTestBase.java
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/BalancerTestBase.java
index 5e4e714..669f1ca 100644
---
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/BalancerTestBase.java
+++
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/BalancerTestBase.java
@@ -75,7 +75,7 @@ public class BalancerTestBase {
conf.setFloat("hbase.master.balancer.stochastic.maxMovePercent", 0.75f);
conf.setFloat("hbase.regions.slop", 0.0f);
conf.setFloat("hbase.master.balancer.stochastic.localityCost", 0);
- conf.setLong(StochasticLoadBalancer.MAX_RUNNING_TIME_KEY, 3 * 60 * 1000L);
+ conf.setBoolean("hbase.master.balancer.stochastic.runMaxSteps", true);
loadBalancer = new StochasticLoadBalancer();
MasterServices services = mock(MasterServices.class);
when(services.getConfiguration()).thenReturn(conf);
diff --git
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestBaseLoadBalancer.java
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestBaseLoadBalancer.java
index 9b81c0d..9d6c4fb 100644
---
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestBaseLoadBalancer.java
+++
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestBaseLoadBalancer.java
@@ -369,8 +369,8 @@ public class TestBaseLoadBalancer extends BalancerTestBase {
// now move region1 from servers[0] to servers[2]
cluster.doAction(new MoveRegionAction(0, 0, 2));
- // check that the numMaxRegionsPerTable for "table" has increased to 2
- assertEquals(2, cluster.numMaxRegionsPerTable[0]);
+ // check that the regionSkewByTable for "table" has increased to 2
+ assertEquals(2, cluster.regionSkewByTable[0], 0.01);
// now repeat check whether moving region1 from servers[1] to servers[2]
// would lower availability
assertTrue(cluster.wouldLowerAvailability(hri1, servers[2]));
diff --git
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerBalanceCluster.java
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerBalanceCluster.java
index d1b597b..eb657d2 100644
---
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerBalanceCluster.java
+++
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerBalanceCluster.java
@@ -51,8 +51,7 @@ public class TestStochasticLoadBalancerBalanceCluster extends
BalancerTestBase {
*/
@Test
public void testBalanceCluster() throws Exception {
- conf.setLong(StochasticLoadBalancer.MAX_STEPS_KEY, 2000000L);
- conf.setLong("hbase.master.balancer.stochastic.maxRunningTime", 90 *
1000); // 90 sec
+ conf.setLong("hbase.master.balancer.stochastic.maxRunningTime", 3 * 60 *
1000); // 3 min
conf.setFloat("hbase.master.balancer.stochastic.maxMovePercent", 1.0f);
loadBalancer.onConfigurationChange(conf);
for (int[] mockCluster : clusterStateMocks) {
diff --git
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerHeterogeneousCost.java
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerHeterogeneousCost.java
index 7d70318..cd6de3f 100644
---
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerHeterogeneousCost.java
+++
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerHeterogeneousCost.java
@@ -67,7 +67,6 @@ public class TestStochasticLoadBalancerHeterogeneousCost
extends BalancerTestBas
BalancerTestBase.conf.setFloat("hbase.master.balancer.stochastic.regionCountCost",
0);
BalancerTestBase.conf.setFloat("hbase.master.balancer.stochastic.primaryRegionCountCost",
0);
BalancerTestBase.conf.setFloat("hbase.master.balancer.stochastic.tableSkewCost",
0);
-
BalancerTestBase.conf.setBoolean("hbase.master.balancer.stochastic.runMaxSteps",
true);
BalancerTestBase.conf.set(StochasticLoadBalancer.COST_FUNCTIONS_COST_FUNCTIONS_KEY,
HeterogeneousRegionCountCostFunction.class.getName());
// Need to ensure test dir has been created.
diff --git
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerLargeCluster.java
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerLargeCluster.java
index da38187..e31cf13 100644
---
a/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerLargeCluster.java
+++
b/hbase-server/src/test/java/org/apache/hadoop/hbase/master/balancer/TestStochasticLoadBalancerLargeCluster.java
@@ -38,6 +38,9 @@ public class TestStochasticLoadBalancerLargeCluster extends
BalancerTestBase {
int numRegionsPerServer = 80; // all servers except one
int numTables = 100;
int replication = 1;
+ conf.setLong("hbase.master.balancer.stochastic.maxRunningTime", 6 * 60 *
1000);
+ conf.setFloat("hbase.master.balancer.stochastic.maxMovePercent", 1.0f);
+ loadBalancer.onConfigurationChange(conf);
testWithCluster(numNodes, numRegions, numRegionsPerServer, replication,
numTables, true, true);
}
}
