Skip to content

Commit 7fcdb36

Browse files
carlosdelestcarlosdelest
authored
Merge branch 'main' into non-issue/esql-vector-search-functions-basics
2 parents 707b3c2 + e4ff3f8 commit 7fcdb36

File tree

515 files changed

+15093
-7764
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

515 files changed

+15093
-7764
lines changed

‎README.asciidoc‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -275,7 +275,7 @@ For the complete Elasticsearch documentation visit
275275
https://www.elastic.co/guide/en/elasticsearch/reference/current/index.html[elastic.co].
276276

277277
For information about our documentation processes, see the
278-
xref:docs/README.asciidoc[docs README].
278+
xref:https://github.com/elastic/elasticsearch/blob/main/docs/README.md[docs README].
279279

280280
[[examples]]
281281
== Examples and guides

‎benchmarks/src/main/java/org/elasticsearch/benchmark/_nightly/esql/QueryPlanningBenchmark.java‎

Lines changed: 3 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -70,11 +70,11 @@ public class QueryPlanningBenchmark {
7070
private EsqlParser defaultParser;
7171
private Analyzer manyFieldsAnalyzer;
7272
private LogicalPlanOptimizer defaultOptimizer;
73+
private Configuration config;
7374

7475
@Setup
7576
public void setup() {
76-
77-
var config = new Configuration(
77+
this.config = new Configuration(
7878
DateUtils.UTC,
7979
Locale.US,
8080
null,
@@ -116,7 +116,7 @@ public void setup() {
116116
}
117117

118118
private LogicalPlan plan(EsqlParser parser, Analyzer analyzer, LogicalPlanOptimizer optimizer, String query) {
119-
var parsed = parser.createStatement(query, new QueryParams(), telemetry);
119+
var parsed = parser.createStatement(query, new QueryParams(), telemetry, config);
120120
var analyzed = analyzer.analyze(parsed);
121121
var optimized = optimizer.optimize(analyzed);
122122
return optimized;

‎benchmarks/src/main/java/org/elasticsearch/benchmark/compute/operator/ValuesSourceReaderBenchmark.java‎ renamed to ‎benchmarks/src/main/java/org/elasticsearch/benchmark/_nightly/esql/ValuesSourceReaderBenchmark.java‎

Lines changed: 15 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,7 @@
77
* License v3.0 only", or the "Server Side Public License, v 1".
88
*/
99

10-
package org.elasticsearch.benchmark.compute.operator;
10+
package org.elasticsearch.benchmark._nightly.esql;
1111

1212
import org.apache.lucene.document.FieldType;
1313
import org.apache.lucene.document.NumericDocValuesField;
@@ -85,10 +85,18 @@
8585
@State(Scope.Thread)
8686
@Fork(1)
8787
public class ValuesSourceReaderBenchmark {
88+
private static final String[] SUPPORTED_LAYOUTS = new String[] { "in_order", "shuffled", "shuffled_singles" };
89+
private static final String[] SUPPORTED_NAMES = new String[] {
90+
"long",
91+
"int",
92+
"double",
93+
"keyword",
94+
"stored_keyword",
95+
"3_stored_keywords" };
96+
8897
private static final int BLOCK_LENGTH = 16 * 1024;
8998
private static final int INDEX_SIZE = 10 * BLOCK_LENGTH;
9099
private static final int COMMIT_INTERVAL = 500;
91-
private static final BigArrays BIG_ARRAYS = BigArrays.NON_RECYCLING_INSTANCE;
92100
private static final BlockFactory blockFactory = BlockFactory.getInstance(
93101
new NoopCircuitBreaker("noop"),
94102
BigArrays.NON_RECYCLING_INSTANCE
@@ -104,8 +112,8 @@ static void selfTest() {
104112
ValuesSourceReaderBenchmark benchmark = new ValuesSourceReaderBenchmark();
105113
benchmark.setupIndex();
106114
try {
107-
for (String layout : ValuesSourceReaderBenchmark.class.getField("layout").getAnnotationsByType(Param.class)[0].value()) {
108-
for (String name : ValuesSourceReaderBenchmark.class.getField("name").getAnnotationsByType(Param.class)[0].value()) {
115+
for (String layout : ValuesSourceReaderBenchmark.SUPPORTED_LAYOUTS) {
116+
for (String name : ValuesSourceReaderBenchmark.SUPPORTED_NAMES) {
109117
benchmark.layout = layout;
110118
benchmark.name = name;
111119
try {
@@ -119,7 +127,7 @@ static void selfTest() {
119127
} finally {
120128
benchmark.teardownIndex();
121129
}
122-
} catch (IOException | NoSuchFieldException e) {
130+
} catch (IOException e) {
123131
throw new AssertionError(e);
124132
}
125133
}
@@ -321,10 +329,10 @@ public FieldNamesFieldMapper.FieldNamesFieldType fieldNames() {
321329
* each page has a single document rather than {@code BLOCK_SIZE} docs.</li>
322330
* </ul>
323331
*/
324-
@Param({ "in_order", "shuffled", "shuffled_singles" })
332+
@Param({ "in_order", "shuffled" })
325333
public String layout;
326334

327-
@Param({ "long", "int", "double", "keyword", "stored_keyword", "3_stored_keywords" })
335+
@Param({ "long", "keyword", "stored_keyword" })
328336
public String name;
329337

330338
private Directory directory;

‎benchmarks/src/main/java/org/elasticsearch/benchmark/vector/Int4ScorerBenchmark.java‎

Lines changed: 72 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -8,12 +8,14 @@
88
*/
99
package org.elasticsearch.benchmark.vector;
1010

11+
import org.apache.lucene.index.VectorSimilarityFunction;
1112
import org.apache.lucene.store.Directory;
1213
import org.apache.lucene.store.IOContext;
1314
import org.apache.lucene.store.IndexInput;
1415
import org.apache.lucene.store.IndexOutput;
1516
import org.apache.lucene.store.MMapDirectory;
1617
import org.apache.lucene.util.VectorUtil;
18+
import org.apache.lucene.util.quantization.OptimizedScalarQuantizer;
1719
import org.elasticsearch.common.logging.LogConfigurator;
1820
import org.elasticsearch.core.IOUtils;
1921
import org.elasticsearch.simdvec.ES91Int4VectorsScorer;
@@ -52,20 +54,26 @@ public class Int4ScorerBenchmark {
5254
LogConfigurator.configureESLogging(); // native access requires logging to be initialized
5355
}
5456

55-
@Param({ "384", "702", "1024" })
57+
@Param({ "384", "782", "1024" })
5658
int dims;
5759

58-
int numVectors = 200;
59-
int numQueries = 10;
60+
int numVectors = 20 * ES91Int4VectorsScorer.BULK_SIZE;
61+
int numQueries = 5;
6062

6163
byte[] scratch;
6264
byte[][] binaryVectors;
6365
byte[][] binaryQueries;
66+
float[] scores = new float[ES91Int4VectorsScorer.BULK_SIZE];
67+
68+
float[] scratchFloats = new float[3];
6469

6570
ES91Int4VectorsScorer scorer;
6671
Directory dir;
6772
IndexInput in;
6873

74+
OptimizedScalarQuantizer.QuantizationResult queryCorrections;
75+
float centroidDp;
76+
6977
@Setup
7078
public void setup() throws IOException {
7179
binaryVectors = new byte[numVectors][dims];
@@ -77,9 +85,19 @@ public void setup() throws IOException {
7785
binaryVector[i] = (byte) ThreadLocalRandom.current().nextInt(16);
7886
}
7987
out.writeBytes(binaryVector, 0, binaryVector.length);
88+
ThreadLocalRandom.current().nextBytes(binaryVector);
89+
out.writeBytes(binaryVector, 0, 14); // corrections
8090
}
8191
}
8292

93+
queryCorrections = new OptimizedScalarQuantizer.QuantizationResult(
94+
ThreadLocalRandom.current().nextFloat(),
95+
ThreadLocalRandom.current().nextFloat(),
96+
ThreadLocalRandom.current().nextFloat(),
97+
Short.toUnsignedInt((short) ThreadLocalRandom.current().nextInt())
98+
);
99+
centroidDp = ThreadLocalRandom.current().nextFloat();
100+
83101
in = dir.openInput("vectors", IOContext.DEFAULT);
84102
binaryQueries = new byte[numVectors][dims];
85103
for (byte[] binaryVector : binaryVectors) {
@@ -105,18 +123,66 @@ public void scoreFromArray(Blackhole bh) throws IOException {
105123
in.seek(0);
106124
for (int i = 0; i < numVectors; i++) {
107125
in.readBytes(scratch, 0, dims);
108-
bh.consume(VectorUtil.int4DotProduct(binaryQueries[j], scratch));
126+
int dp = VectorUtil.int4DotProduct(binaryQueries[j], scratch);
127+
in.readFloats(scratchFloats, 0, 3);
128+
float score = scorer.applyCorrections(
129+
queryCorrections.lowerInterval(),
130+
queryCorrections.upperInterval(),
131+
queryCorrections.quantizedComponentSum(),
132+
queryCorrections.additionalCorrection(),
133+
VectorSimilarityFunction.EUCLIDEAN,
134+
centroidDp, // assuming no centroid dot product for this benchmark
135+
scratchFloats[0],
136+
scratchFloats[1],
137+
Short.toUnsignedInt(in.readShort()),
138+
scratchFloats[2],
139+
dp
140+
);
141+
bh.consume(score);
109142
}
110143
}
111144
}
112145

113146
@Benchmark
114147
@Fork(jvmArgsPrepend = { "--add-modules=jdk.incubator.vector" })
115-
public void scoreFromMemorySegmentOnlyVector(Blackhole bh) throws IOException {
148+
public void scoreFromMemorySegment(Blackhole bh) throws IOException {
116149
for (int j = 0; j < numQueries; j++) {
117150
in.seek(0);
118151
for (int i = 0; i < numVectors; i++) {
119-
bh.consume(scorer.int4DotProduct(binaryQueries[j]));
152+
bh.consume(
153+
scorer.score(
154+
binaryQueries[j],
155+
queryCorrections.lowerInterval(),
156+
queryCorrections.upperInterval(),
157+
queryCorrections.quantizedComponentSum(),
158+
queryCorrections.additionalCorrection(),
159+
VectorSimilarityFunction.EUCLIDEAN,
160+
centroidDp
161+
)
162+
);
163+
}
164+
}
165+
}
166+
167+
@Benchmark
168+
@Fork(jvmArgsPrepend = { "--add-modules=jdk.incubator.vector" })
169+
public void scoreFromMemorySegmentBulk(Blackhole bh) throws IOException {
170+
for (int j = 0; j < numQueries; j++) {
171+
in.seek(0);
172+
for (int i = 0; i < numVectors; i += ES91Int4VectorsScorer.BULK_SIZE) {
173+
scorer.scoreBulk(
174+
binaryQueries[j],
175+
queryCorrections.lowerInterval(),
176+
queryCorrections.upperInterval(),
177+
queryCorrections.quantizedComponentSum(),
178+
queryCorrections.additionalCorrection(),
179+
VectorSimilarityFunction.EUCLIDEAN,
180+
centroidDp,
181+
scores
182+
);
183+
for (float score : scores) {
184+
bh.consume(score);
185+
}
120186
}
121187
}
122188
}

‎benchmarks/src/test/java/org/elasticsearch/benchmark/compute/operator/ValuesSourceReaderBenchmarkTests.java‎ renamed to ‎benchmarks/src/test/java/org/elasticsearch/benchmark/_nightly/esql/ValuesSourceReaderBenchmarkTests.java‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -7,7 +7,7 @@
77
* License v3.0 only", or the "Server Side Public License, v 1".
88
*/
99

10-
package org.elasticsearch.benchmark.compute.operator;
10+
package org.elasticsearch.benchmark._nightly.esql;
1111

1212
import org.elasticsearch.test.ESTestCase;
1313

‎build-tools/src/main/java/org/elasticsearch/gradle/testclusters/RunTask.java‎

Lines changed: 0 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -303,7 +303,6 @@ else if (node.getSettingKeys().contains("telemetry.metrics.enabled") == false) {
303303
if (cliDebug) {
304304
enableCliDebug();
305305
}
306-
enableEntitlements();
307306
}
308307

309308
@TaskAction

‎build-tools/src/main/java/org/elasticsearch/gradle/testclusters/TestClustersAware.java‎

Lines changed: 0 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -88,12 +88,4 @@ default void enableCliDebug() {
8888
}
8989
}
9090
}
91-
92-
default void enableEntitlements() {
93-
for (ElasticsearchCluster cluster : getClusters()) {
94-
for (ElasticsearchNode node : cluster.getNodes()) {
95-
node.cliJvmArgs("-Des.entitlements.enabled=true");
96-
}
97-
}
98-
}
9991
}

‎distribution/tools/server-cli/src/main/java/org/elasticsearch/server/cli/MachineDependentHeap.java‎

Lines changed: 8 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -40,6 +40,8 @@ public class MachineDependentHeap {
4040

4141
private static final FeatureFlag NEW_ML_MEMORY_COMPUTATION_FEATURE_FLAG = new FeatureFlag("new_ml_memory_computation");
4242

43+
private boolean useNewMlMemoryComputation = false;
44+
4345
public MachineDependentHeap() {}
4446

4547
/**
@@ -55,6 +57,11 @@ public final List<String> determineHeapSettings(
5557
SystemMemoryInfo systemMemoryInfo,
5658
List<String> userDefinedJvmOptions
5759
) throws IOException, InterruptedException {
60+
if (userDefinedJvmOptions.contains("-Des.new_ml_memory_computation_feature_flag_enabled=true")
61+
|| NEW_ML_MEMORY_COMPUTATION_FEATURE_FLAG.isEnabled()) {
62+
useNewMlMemoryComputation = true;
63+
}
64+
5865
// TODO: this could be more efficient, to only parse final options once
5966
final Map<String, JvmOption> finalJvmOptions = JvmOption.findFinalOptions(userDefinedJvmOptions);
6067
if (isMaxHeapSpecified(finalJvmOptions) || isMinHeapSpecified(finalJvmOptions) || isInitialHeapSpecified(finalJvmOptions)) {
@@ -107,7 +114,7 @@ protected int getHeapSizeMb(Settings nodeSettings, MachineNodeRole role, long av
107114
case ML_ONLY -> {
108115
double heapFractionBelow16GB = 0.4;
109116
double heapFractionAbove16GB = 0.1;
110-
if (NEW_ML_MEMORY_COMPUTATION_FEATURE_FLAG.isEnabled()) {
117+
if (useNewMlMemoryComputation) {
111118
heapFractionBelow16GB = 0.4 / (1.0 + JvmErgonomics.DIRECT_MEMORY_TO_HEAP_FACTOR);
112119
heapFractionAbove16GB = 0.1 / (1.0 + JvmErgonomics.DIRECT_MEMORY_TO_HEAP_FACTOR);
113120
}

‎distribution/tools/server-cli/src/main/java/org/elasticsearch/server/cli/SystemJvmOptions.java‎

Lines changed: 0 additions & 11 deletions
Original file line numberDiff line numberDiff line change
@@ -11,7 +11,6 @@
1111

1212
import org.elasticsearch.common.settings.Settings;
1313
import org.elasticsearch.common.util.concurrent.EsExecutors;
14-
import org.elasticsearch.jdk.RuntimeVersionFeature;
1514

1615
import java.io.IOException;
1716
import java.nio.file.Files;
@@ -85,7 +84,6 @@ static List<String> systemJvmOptions(Settings nodeSettings, final Map<String, St
8584
maybeSetActiveProcessorCount(nodeSettings),
8685
maybeSetReplayFile(distroType, isHotspot),
8786
maybeWorkaroundG1Bug(),
88-
maybeAllowSecurityManager(useEntitlements),
8987
maybeAttachEntitlementAgent(esHome, useEntitlements)
9088
).flatMap(s -> s).toList();
9189
}
@@ -160,14 +158,6 @@ private static Stream<String> maybeWorkaroundG1Bug() {
160158
return Stream.of();
161159
}
162160

163-
private static Stream<String> maybeAllowSecurityManager(boolean useEntitlements) {
164-
if (RuntimeVersionFeature.isSecurityManagerAvailable()) {
165-
// Will become conditional on useEntitlements once entitlements can run without SM
166-
return Stream.of("-Djava.security.manager=allow");
167-
}
168-
return Stream.of();
169-
}
170-
171161
private static Stream<String> maybeAttachEntitlementAgent(Path esHome, boolean useEntitlements) {
172162
if (useEntitlements == false) {
173163
return Stream.empty();
@@ -191,7 +181,6 @@ private static Stream<String> maybeAttachEntitlementAgent(Path esHome, boolean u
191181
// into java.base, we must export the bridge from java.base to these modules, as a comma-separated list
192182
String modulesContainingEntitlementInstrumentation = "java.logging,java.net.http,java.naming,jdk.net";
193183
return Stream.of(
194-
"-Des.entitlements.enabled=true",
195184
"-XX:+EnableDynamicAgentLoading",
196185
"-Djdk.attach.allowAttachSelf=true",
197186
"--patch-module=java.base=" + bridgeJar,

‎docs/changelog/129013.yaml‎

Lines changed: 13 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,13 @@
1+
pr: 129013
2+
summary: "Add remote index support to LOOKUP JOIN"
3+
area: ES|QL
4+
type: feature
5+
issues: [ ]
6+
highlight:
7+
title: Add remote index support to LOOKUP JOIN
8+
body: |-
9+
Queries containing LOOKUP JOIN now can be preformed on cross-cluster indices, for example:
10+
[source,yaml]
11+
----------------------------
12+
FROM logs-*, remote:logs-* | LOOKUP JOIN clients on ip | SORT timestamp | LIMIT 100
13+
----------------------------

0 commit comments

Comments
 (0)