Skip to content

Commit

Permalink
[Segment Replication] Introduce primary weight factor for primary sha…
Browse files Browse the repository at this point in the history 
…rds distribution (#6017) (#6161)

* Add integration test to show shard allocation



* Update WeightFunction to consider primary shards for uniform primary distribution



* Include primary shard weight for all shard types



* Update integration test to show docrep & segrep indices



* Add settings updater and update TriConsumer functional interface declaration



* Add balance configuration test



* Fix failing unit tests and merge conflicts



* Rename primary balance factor



* Update integration tests to avoid green state timeouts



* PR feedback



---------


(cherry picked from commit aa21b5f)

Signed-off-by: Suraj Singh <surajrider@gmail.com>
Signed-off-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
Co-authored-by: github-actions[bot] <github-actions[bot]@users.noreply.github.com>
  • Loading branch information
@opensearch-trigger-bot @github-actions
opensearch-trigger-bot[bot] and github-actions[bot] committed Feb 2, 2023
1 parent 0963950 commit 2e3e1ed
Show file tree
Hide file tree
Showing 15 changed files with 852 additions and 24 deletions.
63 changes: 63 additions & 0 deletions libs/common/src/main/java/org/opensearch/common/collect/Triplet.java
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,63 @@
/*
* SPDX-License-Identifier: Apache-2.0
*
* The OpenSearch Contributors require contributions made to
* this file be licensed under the Apache-2.0 license or a
* compatible open source license.
*/
package org.opensearch.common.collect;

import java.util.Objects;

/**
* A container for 3 elements, similar to {@link org.opensearch.common.collect.Tuple}
*
* @opensearch.internal
*/
public class Triplet<V1, V2, V3> {

public static <V1, V2, V3> Triplet<V1, V2, V3> tuple(V1 v1, V2 v2, V3 v3) {
return new Triplet<>(v1, v2, v3);
}

private final V1 v1;
private final V2 v2;

private final V3 v3;

@Override
public boolean equals(Object o) {
if (this == o) return true;
if (o == null || getClass() != o.getClass()) return false;
Triplet<?, ?, ?> triplet = (Triplet<?, ?, ?>) o;
return Objects.equals(v1, triplet.v1) && Objects.equals(v2, triplet.v2) && Objects.equals(v3, triplet.v3);
}

@Override
public int hashCode() {
return Objects.hash(v1, v2, v3);
}

public Triplet(V1 v1, V2 v2, V3 v3) {
this.v1 = v1;
this.v2 = v2;
this.v3 = v3;
}

public V1 v1() {
return v1;
}

public V2 v2() {
return v2;
}

public V3 v3() {
return v3;
}

@Override
public String toString() {
return "Tuple [v1=" + v1 + ", v2=" + v2 + ", v3=" + v3 + "]";
}
}
307 changes: 307 additions & 0 deletions ...alClusterTest/java/org/opensearch/indices/replication/SegmentReplicationAllocationIT.java
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,307 @@
/*
* SPDX-License-Identifier: Apache-2.0
*
* The OpenSearch Contributors require contributions made to
* this file be licensed under the Apache-2.0 license or a
* compatible open source license.
*/

package org.opensearch.indices.replication;

import org.opensearch.cluster.ClusterState;
import org.opensearch.cluster.metadata.IndexMetadata;
import org.opensearch.cluster.routing.RoutingNode;
import org.opensearch.cluster.routing.RoutingNodes;
import org.opensearch.cluster.routing.ShardRouting;
import org.opensearch.cluster.routing.allocation.allocator.BalancedShardsAllocator;
import org.opensearch.common.settings.Settings;
import org.opensearch.index.IndexModule;
import org.opensearch.indices.replication.common.ReplicationType;
import org.opensearch.test.InternalTestCluster;
import org.opensearch.test.OpenSearchIntegTestCase;

import java.util.ArrayList;
import java.util.Formatter;
import java.util.List;
import java.util.Locale;
import java.util.Map;
import java.util.TreeMap;
import java.util.concurrent.TimeUnit;

import static org.opensearch.cluster.routing.ShardRoutingState.STARTED;
import static org.opensearch.test.hamcrest.OpenSearchAssertions.assertAcked;

@OpenSearchIntegTestCase.ClusterScope(scope = OpenSearchIntegTestCase.Scope.TEST, numDataNodes = 0)
public class SegmentReplicationAllocationIT extends SegmentReplicationBaseIT {

private void createIndex(String idxName, int shardCount, int replicaCount, boolean isSegRep) {
Settings.Builder builder = Settings.builder()
.put(IndexMetadata.SETTING_NUMBER_OF_SHARDS, shardCount)
.put(IndexModule.INDEX_QUERY_CACHE_ENABLED_SETTING.getKey(), false)
.put(IndexMetadata.SETTING_NUMBER_OF_REPLICAS, REPLICA_COUNT);
if (isSegRep) {
builder = builder.put(IndexMetadata.SETTING_REPLICATION_TYPE, ReplicationType.SEGMENT);
} else {
builder = builder.put(IndexMetadata.SETTING_REPLICATION_TYPE, ReplicationType.DOCUMENT);
}
prepareCreate(idxName, builder).get();
}

/**
* This test verifies primary shard allocation is balanced.
*/
public void testShardAllocation() throws Exception {
internalCluster().startClusterManagerOnlyNode();
final int maxReplicaCount = 2;
final int maxShardCount = 5;
final int nodeCount = randomIntBetween(maxReplicaCount + 1, 10);
final int numberOfIndices = randomIntBetween(5, 10);

final List<String> nodeNames = new ArrayList<>();
logger.info("--> Creating {} nodes", nodeCount);
for (int i = 0; i < nodeCount; i++) {
nodeNames.add(internalCluster().startNode());
}
assertAcked(
client().admin()
.cluster()
.prepareUpdateSettings()
.setPersistentSettings(
Settings.builder().put(BalancedShardsAllocator.PRIMARY_SHARD_BALANCE_FACTOR_SETTING.getKey(), "1.0f")
)
);

int shardCount, replicaCount, totalShardCount = 0, totalReplicaCount = 0;
ShardAllocations shardAllocations = new ShardAllocations();
ClusterState state;
for (int i = 0; i < numberOfIndices; i++) {
shardCount = randomIntBetween(1, maxShardCount);
totalShardCount += shardCount;
replicaCount = randomIntBetween(0, maxReplicaCount);
totalReplicaCount += replicaCount;
createIndex("test" + i, shardCount, replicaCount, i % 2 == 0);
logger.info("--> Creating index {} with shard count {} and replica count {}", "test" + i, shardCount, replicaCount);
assertBusy(() -> ensureGreen(), 60, TimeUnit.SECONDS);
state = client().admin().cluster().prepareState().execute().actionGet().getState();
shardAllocations.printShardDistribution(state);
}
state = client().admin().cluster().prepareState().execute().actionGet().getState();
RoutingNodes nodes = state.getRoutingNodes();
final float avgNumShards = (float) (totalShardCount) / (float) (nodes.size());
final int minAvgNumberOfShards = Math.round(Math.round(Math.floor(avgNumShards - 1.0f)));
final int maxAvgNumberOfShards = Math.round(Math.round(Math.ceil(avgNumShards + 1.0f)));

for (RoutingNode node : nodes) {
assertTrue(node.primaryShardsWithState(STARTED).size() >= minAvgNumberOfShards);
assertTrue(node.primaryShardsWithState(STARTED).size() <= maxAvgNumberOfShards);
}
}

/**
* This test verifies shard allocation with changes to cluster config i.e. node add, removal keeps the primary shard
* allocation balanced.
*/
public void testAllocationWithDisruption() throws Exception {
internalCluster().startClusterManagerOnlyNode();
final int maxReplicaCount = 2;
final int maxShardCount = 5;
final int nodeCount = randomIntBetween(maxReplicaCount + 1, 10);
final int numberOfIndices = randomIntBetween(1, 10);

logger.info("--> Creating {} nodes", nodeCount);
final List<String> nodeNames = new ArrayList<>();
for (int i = 0; i < nodeCount; i++) {
nodeNames.add(internalCluster().startNode());
}
assertAcked(
client().admin()
.cluster()
.prepareUpdateSettings()
.setPersistentSettings(
Settings.builder()
.put(BalancedShardsAllocator.PRIMARY_SHARD_BALANCE_FACTOR_SETTING.getKey(), "1.0f")
.put(BalancedShardsAllocator.INDEX_BALANCE_FACTOR_SETTING.getKey(), "0.0f")
.put(BalancedShardsAllocator.SHARD_BALANCE_FACTOR_SETTING.getKey(), "0.0f")
.build()
)
);

int shardCount, replicaCount, totalShardCount = 0, totalReplicaCount = 0;
ShardAllocations shardAllocations = new ShardAllocations();
ClusterState state;
for (int i = 0; i < numberOfIndices; i++) {
shardCount = randomIntBetween(1, maxShardCount);
totalShardCount += shardCount;
replicaCount = randomIntBetween(1, maxReplicaCount);
totalReplicaCount += replicaCount;
logger.info("--> Creating index test{} with primary {} and replica {}", i, shardCount, replicaCount);
createIndex("test" + i, shardCount, replicaCount, i % 2 == 0);
assertBusy(() -> ensureGreen(), 60, TimeUnit.SECONDS);
if (logger.isTraceEnabled()) {
state = client().admin().cluster().prepareState().execute().actionGet().getState();
shardAllocations.printShardDistribution(state);
}
}
state = client().admin().cluster().prepareState().execute().actionGet().getState();
float avgNumShards = (float) (totalShardCount) / (float) (state.getRoutingNodes().size());
int minAvgNumberOfShards = Math.round(Math.round(Math.floor(avgNumShards - 1.0f)));
int maxAvgNumberOfShards = Math.round(Math.round(Math.ceil(avgNumShards + 1.0f)));

for (RoutingNode node : state.getRoutingNodes()) {
assertTrue(node.primaryShardsWithState(STARTED).size() >= minAvgNumberOfShards);
assertTrue(node.primaryShardsWithState(STARTED).size() <= maxAvgNumberOfShards);
}

final int additionalNodeCount = randomIntBetween(1, 5);
logger.info("--> Adding {} nodes", additionalNodeCount);

internalCluster().startNodes(additionalNodeCount);
assertBusy(() -> ensureGreen(), 60, TimeUnit.SECONDS);
state = client().admin().cluster().prepareState().execute().actionGet().getState();
avgNumShards = (float) (totalShardCount) / (float) (state.getRoutingNodes().size());
minAvgNumberOfShards = Math.round(Math.round(Math.floor(avgNumShards - 1.0f)));
maxAvgNumberOfShards = Math.round(Math.round(Math.ceil(avgNumShards + 1.0f)));
shardAllocations.printShardDistribution(state);
for (RoutingNode node : state.getRoutingNodes()) {
assertTrue(node.primaryShardsWithState(STARTED).size() >= minAvgNumberOfShards);
assertTrue(node.primaryShardsWithState(STARTED).size() <= maxAvgNumberOfShards);
}

logger.info("--> Stop one third nodes");
for (int i = 1; i < nodeCount; i += 3) {
internalCluster().stopRandomNode(InternalTestCluster.nameFilter(nodeNames.get(i)));
// give replica a chance to promote as primary before terminating node containing the replica
assertBusy(() -> ensureGreen(), 60, TimeUnit.SECONDS);
}
state = client().admin().cluster().prepareState().execute().actionGet().getState();
avgNumShards = (float) (totalShardCount) / (float) (state.getRoutingNodes().size());
minAvgNumberOfShards = Math.round(Math.round(Math.floor(avgNumShards - 1.0f)));
maxAvgNumberOfShards = Math.round(Math.round(Math.ceil(avgNumShards + 1.0f)));
shardAllocations.printShardDistribution(state);

for (RoutingNode node : state.getRoutingNodes()) {
assertTrue(node.primaryShardsWithState(STARTED).size() >= minAvgNumberOfShards);
assertTrue(node.primaryShardsWithState(STARTED).size() <= maxAvgNumberOfShards);
}
}

/**
* This class is created for debugging purpose to show shard allocation across nodes. It keeps cluster state which
* is used to build the node's shard allocation
*/
private class ShardAllocations {
ClusterState state;

public static final String separator = "===================================================";
public static final String ONE_LINE_RETURN = "\n";
public static final String TWO_LINE_RETURN = "\n\n";

/**
Store shard primary/replica shard count against a node for segrep indices.
String: NodeId
int[]: tuple storing primary shard count in 0th index and replica's in 1
*/
TreeMap<String, int[]> nodeToSegRepCountMap = new TreeMap<>();
/**
Store shard primary/replica shard count against a node for docrep indices.
String: NodeId
int[]: tuple storing primary shard count in 0th index and replica's in 1
*/
TreeMap<String, int[]> nodeToDocRepCountMap = new TreeMap<>();

/**
* Helper map containing NodeName to NodeId
*/
TreeMap<String, String> nameToNodeId = new TreeMap<>();

/*
Unassigned array containing primary at 0, replica at 1
*/
int[] unassigned = new int[2];

int[] totalShards = new int[2];

public final String printShardAllocationWithHeader(int[] docrep, int[] segrep) {
StringBuffer sb = new StringBuffer();
Formatter formatter = new Formatter(sb, Locale.getDefault());
formatter.format("%-20s %-20s %-20s %-20s\n", "P", docrep[0] + segrep[0], docrep[0], segrep[0]);
formatter.format("%-20s %-20s %-20s %-20s\n", "R", docrep[1] + segrep[1], docrep[1], segrep[1]);
return sb.toString();
}

public void reset() {
nodeToSegRepCountMap.clear();
nodeToDocRepCountMap.clear();
nameToNodeId.clear();
totalShards[0] = totalShards[1] = 0;
unassigned[0] = unassigned[1] = 0;
}

public void setState(ClusterState state) {
this.reset();
this.state = state;
buildMap();
}

private void buildMap() {
for (RoutingNode node : state.getRoutingNodes()) {
nameToNodeId.putIfAbsent(node.node().getName(), node.nodeId());
nodeToSegRepCountMap.putIfAbsent(node.nodeId(), new int[] { 0, 0 });
nodeToDocRepCountMap.putIfAbsent(node.nodeId(), new int[] { 0, 0 });
}
for (ShardRouting shardRouting : state.routingTable().allShards()) {
// Fetch shard to update. Initialize local array
if (isIndexSegRep(shardRouting.getIndexName())) {
updateMap(nodeToSegRepCountMap, shardRouting);
} else {
updateMap(nodeToDocRepCountMap, shardRouting);
}
}
}

void updateMap(TreeMap<String, int[]> mapToUpdate, ShardRouting shardRouting) {
int[] shard;
shard = shardRouting.assignedToNode() ? mapToUpdate.get(shardRouting.currentNodeId()) : unassigned;
// Update shard type count
if (shardRouting.primary()) {
shard[0]++;
totalShards[0]++;
} else {
shard[1]++;
totalShards[1]++;
}
// For assigned shards, put back counter
if (shardRouting.assignedToNode()) mapToUpdate.put(shardRouting.currentNodeId(), shard);
}

boolean isIndexSegRep(String indexName) {
return state.metadata()
.index(indexName)
.getSettings()
.get(IndexMetadata.INDEX_REPLICATION_TYPE_SETTING.getKey())
.equals(ReplicationType.SEGMENT.toString());
}

@Override
public String toString() {
StringBuffer sb = new StringBuffer();
sb.append(TWO_LINE_RETURN + separator + ONE_LINE_RETURN);
Formatter formatter = new Formatter(sb, Locale.getDefault());
for (Map.Entry<String, String> entry : nameToNodeId.entrySet()) {
String nodeId = nameToNodeId.get(entry.getKey());
formatter.format("%-20s %-20s %-20s %-20s\n", entry.getKey().toUpperCase(Locale.getDefault()), "TOTAL", "DOCREP", "SEGREP");
sb.append(printShardAllocationWithHeader(nodeToDocRepCountMap.get(nodeId), nodeToSegRepCountMap.get(nodeId)));
}
sb.append(ONE_LINE_RETURN);
formatter.format("%-20s %-20s %-20s\n\n", "Unassigned ", unassigned[0], unassigned[1]);
formatter.format("%-20s %-20s %-20s\n\n", "Total Shards", totalShards[0], totalShards[1]);
return sb.toString();
}

public void printShardDistribution(ClusterState state) {
this.setState(state);
logger.info("--> Shard distribution {}", this);
}
}

}
17 changes: 17 additions & 0 deletions server/src/main/java/org/opensearch/cluster/routing/RoutingNode.java
View file
Original file line number Diff line number Diff line change
Expand Up @@ -329,6 +329,23 @@ public List<ShardRouting> shardsWithState(ShardRoutingState... states) {
return shards;
}

/**
* Determine the primary shards of an index with a specific state
* @param states set of states which should be listed
* @return a list of shards
*/
public List<ShardRouting> primaryShardsWithState(ShardRoutingState... states) {
List<ShardRouting> shards = new ArrayList<>();
for (ShardRouting shardEntry : this) {
for (ShardRoutingState state : states) {
if (shardEntry.state() == state && shardEntry.primary() == true) {
shards.add(shardEntry);
}
}
}
return shards;
}

/**
* Determine the shards of an index with a specific state
* @param index id of the index
Expand Down
Loading

0 comments on commit 2e3e1ed

Please sign in to comment.