Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Original file line number Diff line number Diff line change
Expand Up @@ -145,7 +145,7 @@ int getViolationCountChange(BalancerClusterState cluster, BalanceAction action)
// Reset cluster
cluster.doAction(undoAction);

if (isViolatingPre && isViolatingPost) {
if (isViolatingPre == isViolatingPost) {
return 0;
Comment on lines -148 to 149
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We also want to return 0 if we had zero violations both pre & post. Prior to this fix, we'd fall through to return -1 which erroneously considered a neutral move to be an improvement

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Since this was so impactful to behavior, maybe good to capture this method in a unit test?

} else if (!isViolatingPre && isViolatingPost) {
return 1;
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -46,16 +46,22 @@ public final class CandidateGeneratorTestUtil {
private CandidateGeneratorTestUtil() {
}

enum ExhaustionType {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

👍

COST_GOAL_ACHIEVED,
NO_MORE_MOVES;
}

static void runBalancerToExhaustion(Configuration conf,
Map<ServerName, List<RegionInfo>> serverToRegions,
Set<Function<BalancerClusterState, Boolean>> expectations, float targetMaxBalancerCost) {
runBalancerToExhaustion(conf, serverToRegions, expectations, targetMaxBalancerCost, 15000);
runBalancerToExhaustion(conf, serverToRegions, expectations, targetMaxBalancerCost, 15000,
ExhaustionType.COST_GOAL_ACHIEVED);
}

static void runBalancerToExhaustion(Configuration conf,
Map<ServerName, List<RegionInfo>> serverToRegions,
Set<Function<BalancerClusterState, Boolean>> expectations, float targetMaxBalancerCost,
long maxRunningTime) {
long maxRunningTime, ExhaustionType exhaustionType) {
// Do the full plan. We're testing with a lot of regions
conf.setBoolean("hbase.master.balancer.stochastic.runMaxSteps", true);
conf.setLong(MAX_RUNNING_TIME_KEY, maxRunningTime);
Expand All @@ -71,7 +77,7 @@ static void runBalancerToExhaustion(Configuration conf,
boolean isBalanced = false;
while (!isBalanced) {
balancerRuns++;
if (balancerRuns > 1000) {
if (balancerRuns > 10) {
Comment on lines -74 to +80
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I validated locally that this change is okay, will see if the tests also pass in CI. If it's taking more than 10 balancer runs for a test, then we should probably increase that tests max run time to make it more efficient

throw new RuntimeException("Balancer failed to find balance & meet expectations");
}
long start = System.currentTimeMillis();
Expand Down Expand Up @@ -106,16 +112,24 @@ static void runBalancerToExhaustion(Configuration conf,
}
}
if (isBalanced) { // Check if the balancer thinks we're done too
LOG.info("All balancer conditions passed. Checking if balancer thinks it's done.");
if (stochasticLoadBalancer.needsBalance(HConstants.ENSEMBLE_TABLE_NAME, cluster)) {
LOG.info("Balancer would still like to run");
isBalanced = false;
if (exhaustionType == ExhaustionType.COST_GOAL_ACHIEVED) {
// If we expect to achieve the cost goal, then needsBalance should be false
if (stochasticLoadBalancer.needsBalance(HConstants.ENSEMBLE_TABLE_NAME, cluster)) {
LOG.info("Balancer cost goal is not achieved. needsBalance=true");
isBalanced = false;
}
} else {
LOG.info("Balancer is done");
// If we anticipate running out of moves, then our last balance run should have produced
// nothing
if (regionPlans != null && !regionPlans.isEmpty()) {
LOG.info("Balancer is not out of moves. regionPlans.size()={}", regionPlans.size());
isBalanced = false;
}
}
}
}
LOG.info("Balancing took {}sec", Duration.ofMillis(balancingMillis).toMinutes());
LOG.info("Balancer is done. Balancing took {}sec",
Duration.ofMillis(balancingMillis).toMinutes());
}

/**
Expand Down
Original file line number Diff line number Diff line change
Expand Up @@ -104,9 +104,10 @@ public void testTableIsolationAndReplicaDistribution() {
conf.setBoolean(BalancerConditionals.ISOLATE_SYSTEM_TABLES_KEY, true);
DistributeReplicasTestConditional.enableConditionalReplicaDistributionForTest(conf);

runBalancerToExhaustion(conf, serverToRegions, Set.of(this::isMetaTableIsolated,
this::isSystemTableIsolated, CandidateGeneratorTestUtil::areAllReplicasDistributed), 10.0f,
60_000);
runBalancerToExhaustion(conf, serverToRegions,
Set.of(this::isMetaTableIsolated, this::isSystemTableIsolated,
CandidateGeneratorTestUtil::areAllReplicasDistributed),
10.0f, 60_000, CandidateGeneratorTestUtil.ExhaustionType.COST_GOAL_ACHIEVED);
LOG.info("Meta table regions are successfully isolated, "
+ "and region replicas are appropriately distributed.");
}
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,108 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one
* or more contributor license agreements. See the NOTICE file
* distributed with this work for additional information
* regarding copyright ownership. The ASF licenses this file
* to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance
* with the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.hadoop.hbase.master.balancer;

import static org.apache.hadoop.hbase.master.balancer.CandidateGeneratorTestUtil.isTableIsolated;
import static org.apache.hadoop.hbase.master.balancer.CandidateGeneratorTestUtil.runBalancerToExhaustion;

import java.util.ArrayList;
import java.util.HashMap;
import java.util.List;
import java.util.Map;
import java.util.Set;
import org.apache.hadoop.conf.Configuration;
import org.apache.hadoop.hbase.HBaseClassTestRule;
import org.apache.hadoop.hbase.ServerName;
import org.apache.hadoop.hbase.TableName;
import org.apache.hadoop.hbase.client.RegionInfo;
import org.apache.hadoop.hbase.client.RegionInfoBuilder;
import org.apache.hadoop.hbase.testclassification.MasterTests;
import org.apache.hadoop.hbase.testclassification.MediumTests;
import org.junit.BeforeClass;
import org.junit.ClassRule;
import org.junit.Test;
import org.junit.experimental.categories.Category;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;

/**
* If your minCostNeedsBalance is set too low, then the balancer should still eventually stop making
* moves as further cost improvements become impossible, and balancer plan calculation becomes
* wasteful. This test ensures that the balancer will not get stuck in a loop of continuously moving
* regions.
*/
@Category({ MasterTests.class, MediumTests.class })
public class TestUnattainableBalancerCostGoal {
Comment on lines +43 to +50
Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This test fails on master atm (if balancer conditionals are enabled). It passes with this patch


@ClassRule
public static final HBaseClassTestRule CLASS_RULE =
HBaseClassTestRule.forClass(TestUnattainableBalancerCostGoal.class);

private static final Logger LOG = LoggerFactory.getLogger(TestUnattainableBalancerCostGoal.class);

private static final TableName SYSTEM_TABLE_NAME = TableName.valueOf("hbase:system");
private static final TableName NON_SYSTEM_TABLE_NAME = TableName.valueOf("userTable");

private static final int NUM_SERVERS = 10;
private static final int NUM_REGIONS = 1000;
private static final float UNACHIEVABLE_COST_GOAL = 0.01f;

private static final ServerName[] servers = new ServerName[NUM_SERVERS];
private static final Map<ServerName, List<RegionInfo>> serverToRegions = new HashMap<>();

@BeforeClass
public static void setup() {
// Initialize servers
for (int i = 0; i < NUM_SERVERS; i++) {
servers[i] = ServerName.valueOf("server" + i, i, System.currentTimeMillis());
}

// Create regions
List<RegionInfo> allRegions = new ArrayList<>();
for (int i = 0; i < NUM_REGIONS; i++) {
TableName tableName = i < 3 ? SYSTEM_TABLE_NAME : NON_SYSTEM_TABLE_NAME;
byte[] startKey = new byte[1];
startKey[0] = (byte) i;
byte[] endKey = new byte[1];
endKey[0] = (byte) (i + 1);

RegionInfo regionInfo =
RegionInfoBuilder.newBuilder(tableName).setStartKey(startKey).setEndKey(endKey).build();
allRegions.add(regionInfo);
}

// Assign all regions to the first server
serverToRegions.put(servers[0], new ArrayList<>(allRegions));
for (int i = 1; i < NUM_SERVERS; i++) {
serverToRegions.put(servers[i], new ArrayList<>());
}
}

@Test
public void testSystemTableIsolation() {
Configuration conf = new Configuration(false);
conf.setBoolean(BalancerConditionals.ISOLATE_SYSTEM_TABLES_KEY, true);
runBalancerToExhaustion(conf, serverToRegions, Set.of(this::isSystemTableIsolated),
UNACHIEVABLE_COST_GOAL, 10_000, CandidateGeneratorTestUtil.ExhaustionType.NO_MORE_MOVES);
LOG.info("Meta table regions are successfully isolated.");
}

private boolean isSystemTableIsolated(BalancerClusterState cluster) {
return isTableIsolated(cluster, SYSTEM_TABLE_NAME, "System");
}
}