Skip to content

Commit bb0a4ae

Browse files
salvatore-campagnasalvatore-campagna
committed
Use synthetic recovery source by default if synthetic source is enabled (elastic#119110)
We experimented with using synthetic source for recovery and observed quite positive impact on indexing throughput by means of our nightly Rally benchmarks. As a result, here we enable it by default when synthetic source is used. To be more precise, if `index.mapping.source.mode` setting is `synthetic` we enable recovery source by means of synthetic source. Moreover, enabling synthetic source recovery is done behind a feature flag. That would allow us to enable it in snapshot builds which in turn will allow us to see performance results in Rally nightly benchmarks. (cherry picked from commit 6a52675)
1 parent 4430854 commit bb0a4ae

File tree

11 files changed

+369
-30
lines changed

11 files changed

+369
-30
lines changed

‎qa/smoke-test-multinode/build.gradle‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -28,5 +28,7 @@ tasks.named("yamlRestTest").configure {
2828
'cat.templates/10_basic/No templates',
2929
'cat.templates/10_basic/Sort templates',
3030
'cat.templates/10_basic/Multiple template',
31+
'update/100_synthetic_source/keyword',
32+
'update/100_synthetic_source/stored text'
3133
].join(',')
3234
}

‎rest-api-spec/src/yamlRestTest/resources/rest-api-spec/test/update/100_synthetic_source.yml‎

Lines changed: 8 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -6,8 +6,8 @@ setup:
66
---
77
keyword:
88
- requires:
9-
cluster_features: ["gte_v8.4.0"]
10-
reason: introduced in 8.4.0
9+
cluster_features: [ "mapper.synthetic_recovery_source" ]
10+
reason: requires synthetic recovery source
1111

1212
- do:
1313
indices.create:
@@ -60,13 +60,14 @@ keyword:
6060
index: test
6161
run_expensive_tasks: true
6262
- is_false: test.fields._source
63-
- is_true: test.fields._recovery_source
63+
# When synthetic source is used there is no _recovery_source field
64+
- match: { test.fields._recovery_source: null }
6465

6566
---
6667
stored text:
6768
- requires:
68-
cluster_features: ["gte_v8.5.0"]
69-
reason: introduced in 8.5.0
69+
cluster_features: [ "mapper.synthetic_recovery_source" ]
70+
reason: requires synthetic recovery source
7071

7172
- do:
7273
indices.create:
@@ -121,4 +122,5 @@ stored text:
121122
index: test
122123
run_expensive_tasks: true
123124
- is_false: test.fields._source
124-
- is_true: test.fields._recovery_source
125+
# When synthetic source is used there is no _recovery_source field
126+
- match: { test.fields._recovery_source: null }

‎server/src/main/java/org/elasticsearch/common/settings/Setting.java‎

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1586,6 +1586,15 @@ public static Setting<Boolean> boolSetting(String key, boolean defaultValue, Val
15861586
return new Setting<>(key, Boolean.toString(defaultValue), b -> parseBoolean(b, key, isFiltered(properties)), validator, properties);
15871587
}
15881588

1589+
public static Setting<Boolean> boolSetting(
1590+
String key,
1591+
Function<Settings, String> defaultValueFn,
1592+
Validator<Boolean> validator,
1593+
Property... properties
1594+
) {
1595+
return new Setting<>(key, defaultValueFn, booleanParser(key, properties), validator, properties);
1596+
}
1597+
15891598
public static Setting<Boolean> boolSetting(String key, Function<Settings, String> defaultValueFn, Property... properties) {
15901599
return new Setting<>(key, defaultValueFn, b -> parseBoolean(b, key, isFiltered(properties)), properties);
15911600
}

‎server/src/main/java/org/elasticsearch/index/IndexSettings.java‎

Lines changed: 15 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -24,6 +24,7 @@
2424
import org.elasticsearch.common.time.DateUtils;
2525
import org.elasticsearch.common.unit.ByteSizeUnit;
2626
import org.elasticsearch.common.unit.ByteSizeValue;
27+
import org.elasticsearch.common.util.FeatureFlag;
2728
import org.elasticsearch.core.TimeValue;
2829
import org.elasticsearch.features.NodeFeature;
2930
import org.elasticsearch.index.mapper.IgnoredSourceFieldMapper;
@@ -40,6 +41,7 @@
4041
import java.util.List;
4142
import java.util.Locale;
4243
import java.util.Map;
44+
import java.util.Objects;
4345
import java.util.concurrent.TimeUnit;
4446
import java.util.function.Consumer;
4547
import java.util.function.Function;
@@ -725,9 +727,19 @@ public Iterator<Setting<?>> settings() {
725727
Setting.Property.IndexScope
726728
);
727729

730+
public static final FeatureFlag RECOVERY_USE_SYNTHETIC_SOURCE = new FeatureFlag("index_recovery_use_synthetic_source");
728731
public static final Setting<Boolean> RECOVERY_USE_SYNTHETIC_SOURCE_SETTING = Setting.boolSetting(
729732
"index.recovery.use_synthetic_source",
730-
false,
733+
settings -> {
734+
boolean isSyntheticSourceRecoveryFeatureFlagEnabled = RECOVERY_USE_SYNTHETIC_SOURCE.isEnabled();
735+
boolean isNewIndexVersion = SETTING_INDEX_VERSION_CREATED.get(settings)
736+
.onOrAfter(IndexVersions.USE_SYNTHETIC_SOURCE_FOR_RECOVERY_BY_DEFAULT_BACKPORT);
737+
boolean useSyntheticRecoverySource = isSyntheticSourceRecoveryFeatureFlagEnabled && isNewIndexVersion;
738+
return String.valueOf(
739+
useSyntheticRecoverySource
740+
&& Objects.equals(INDEX_MAPPER_SOURCE_MODE_SETTING.get(settings), SourceFieldMapper.Mode.SYNTHETIC)
741+
);
742+
},
731743
new Setting.Validator<>() {
732744
@Override
733745
public void validate(Boolean value) {}
@@ -1088,7 +1100,8 @@ public IndexSettings(final IndexMetadata indexMetadata, final Settings nodeSetti
10881100
skipIgnoredSourceRead = scopedSettings.get(IgnoredSourceFieldMapper.SKIP_IGNORED_SOURCE_READ_SETTING);
10891101
indexMappingSourceMode = scopedSettings.get(INDEX_MAPPER_SOURCE_MODE_SETTING);
10901102
recoverySourceEnabled = RecoverySettings.INDICES_RECOVERY_SOURCE_ENABLED_SETTING.get(nodeSettings);
1091-
recoverySourceSyntheticEnabled = scopedSettings.get(RECOVERY_USE_SYNTHETIC_SOURCE_SETTING);
1103+
recoverySourceSyntheticEnabled = DiscoveryNode.isStateless(nodeSettings) == false
1104+
&& scopedSettings.get(RECOVERY_USE_SYNTHETIC_SOURCE_SETTING);
10921105
if (recoverySourceSyntheticEnabled) {
10931106
if (DiscoveryNode.isStateless(settings)) {
10941107
throw new IllegalArgumentException("synthetic recovery source is only allowed in stateful");

‎server/src/main/java/org/elasticsearch/index/IndexVersions.java‎

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -122,10 +122,10 @@ private static IndexVersion def(int id, Version luceneVersion) {
122122
public static final IndexVersion LOGSDB_DEFAULT_IGNORE_DYNAMIC_BEYOND_LIMIT_BACKPORT = def(8_519_0_00, Version.LUCENE_9_12_0);
123123
public static final IndexVersion TIME_BASED_K_ORDERED_DOC_ID_BACKPORT = def(8_520_0_00, Version.LUCENE_9_12_0);
124124
public static final IndexVersion DEPRECATE_SOURCE_MODE_MAPPER = def(8_521_0_00, Version.LUCENE_9_12_0);
125-
public static final IndexVersion USE_SYNTHETIC_SOURCE_FOR_RECOVERY_BACKPORT = def(8_522_0_00, Version.LUCENE_9_12_0);
126125
public static final IndexVersion UPGRADE_TO_LUCENE_9_12_1 = def(8_523_0_00, Version.LUCENE_9_12_1);
127126
public static final IndexVersion INFERENCE_METADATA_FIELDS_BACKPORT = def(8_524_0_00, Version.LUCENE_9_12_1);
128127
public static final IndexVersion LOGSB_OPTIONAL_SORTING_ON_HOST_NAME_BACKPORT = def(8_525_0_00, Version.LUCENE_9_12_1);
128+
public static final IndexVersion USE_SYNTHETIC_SOURCE_FOR_RECOVERY_BACKPORT = def(8_526_0_00, Version.LUCENE_9_12_1);
129129
/*
130130
* STOP! READ THIS FIRST! No, really,
131131
* ____ _____ ___ ____ _ ____ _____ _ ____ _____ _ _ ___ ____ _____ ___ ____ ____ _____ _

‎server/src/test/java/org/elasticsearch/index/mapper/IgnoredSourceFieldMapperTests.java‎

Lines changed: 8 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -2427,8 +2427,14 @@ protected void validateRoundTripReader(String syntheticSource, DirectoryReader r
24272427
// and since the copy is exact, contents of ignored source are different.
24282428
assertReaderEquals(
24292429
"round trip " + syntheticSource,
2430-
new FieldMaskingReader(Set.of(SourceFieldMapper.RECOVERY_SOURCE_NAME, IgnoredSourceFieldMapper.NAME), reader),
2431-
new FieldMaskingReader(Set.of(SourceFieldMapper.RECOVERY_SOURCE_NAME, IgnoredSourceFieldMapper.NAME), roundTripReader)
2430+
new FieldMaskingReader(
2431+
Set.of(SourceFieldMapper.RECOVERY_SOURCE_NAME, IgnoredSourceFieldMapper.NAME, SourceFieldMapper.RECOVERY_SOURCE_SIZE_NAME),
2432+
reader
2433+
),
2434+
new FieldMaskingReader(
2435+
Set.of(SourceFieldMapper.RECOVERY_SOURCE_NAME, IgnoredSourceFieldMapper.NAME, SourceFieldMapper.RECOVERY_SOURCE_SIZE_NAME),
2436+
roundTripReader
2437+
)
24322438
);
24332439
}
24342440
}

‎server/src/test/java/org/elasticsearch/index/mapper/SourceFieldMapperTests.java‎

Lines changed: 47 additions & 16 deletions
Original file line numberDiff line numberDiff line change
@@ -475,8 +475,13 @@ public void testRecoverySourceWithSyntheticSource() throws IOException {
475475
MapperService mapperService = createMapperService(settings, topMapping(b -> {}));
476476
DocumentMapper docMapper = mapperService.documentMapper();
477477
ParsedDocument doc = docMapper.parse(source(b -> b.field("field1", "value1")));
478-
assertNotNull(doc.rootDoc().getField("_recovery_source"));
479-
assertThat(doc.rootDoc().getField("_recovery_source").binaryValue(), equalTo(new BytesRef("{\"field1\":\"value1\"}")));
478+
if (IndexSettings.RECOVERY_USE_SYNTHETIC_SOURCE.isEnabled() == false) {
479+
// TODO: remove this if branch when removing the 'index_recovery_use_synthetic_source' feature flag
480+
assertNotNull(doc.rootDoc().getField("_recovery_source"));
481+
assertThat(doc.rootDoc().getField("_recovery_source").binaryValue(), equalTo(new BytesRef("{\"field1\":\"value1\"}")));
482+
} else {
483+
assertNull(doc.rootDoc().getField("_recovery_source"));
484+
}
480485
}
481486
{
482487
Settings settings = Settings.builder()
@@ -507,8 +512,16 @@ public void testRecoverySourceWithLogs() throws IOException {
507512
MapperService mapperService = createMapperService(settings, mapping(b -> {}));
508513
DocumentMapper docMapper = mapperService.documentMapper();
509514
ParsedDocument doc = docMapper.parse(source(b -> { b.field("@timestamp", "2012-02-13"); }));
510-
assertNotNull(doc.rootDoc().getField("_recovery_source"));
511-
assertThat(doc.rootDoc().getField("_recovery_source").binaryValue(), equalTo(new BytesRef("{\"@timestamp\":\"2012-02-13\"}")));
515+
if (IndexSettings.RECOVERY_USE_SYNTHETIC_SOURCE.isEnabled() == false) {
516+
// TODO: remove this if branch when removing the 'index_recovery_use_synthetic_source' feature flag
517+
assertNotNull(doc.rootDoc().getField("_recovery_source"));
518+
assertThat(
519+
doc.rootDoc().getField("_recovery_source").binaryValue(),
520+
equalTo(new BytesRef("{\"@timestamp\":\"2012-02-13\"}"))
521+
);
522+
} else {
523+
assertNull(doc.rootDoc().getField("_recovery_source"));
524+
}
512525
}
513526
{
514527
Settings settings = Settings.builder()
@@ -701,8 +714,16 @@ public void testRecoverySourceWithLogsCustom() throws IOException {
701714
MapperService mapperService = createMapperService(settings, mappings);
702715
DocumentMapper docMapper = mapperService.documentMapper();
703716
ParsedDocument doc = docMapper.parse(source(b -> { b.field("@timestamp", "2012-02-13"); }));
704-
assertNotNull(doc.rootDoc().getField("_recovery_source"));
705-
assertThat(doc.rootDoc().getField("_recovery_source").binaryValue(), equalTo(new BytesRef("{\"@timestamp\":\"2012-02-13\"}")));
717+
if (IndexSettings.RECOVERY_USE_SYNTHETIC_SOURCE.isEnabled() == false) {
718+
// TODO: remove this if branch when removing the 'index_recovery_use_synthetic_source' feature flag
719+
assertNotNull(doc.rootDoc().getField("_recovery_source"));
720+
assertThat(
721+
doc.rootDoc().getField("_recovery_source").binaryValue(),
722+
equalTo(new BytesRef("{\"@timestamp\":\"2012-02-13\"}"))
723+
);
724+
} else {
725+
assertNull(doc.rootDoc().getField("_recovery_source"));
726+
}
706727
}
707728
{
708729
Settings settings = Settings.builder()
@@ -728,11 +749,16 @@ public void testRecoverySourceWithTimeSeries() throws IOException {
728749
}));
729750
DocumentMapper docMapper = mapperService.documentMapper();
730751
ParsedDocument doc = docMapper.parse(source("123", b -> b.field("@timestamp", "2012-02-13").field("field", "value1"), null));
731-
assertNotNull(doc.rootDoc().getField("_recovery_source"));
732-
assertThat(
733-
doc.rootDoc().getField("_recovery_source").binaryValue(),
734-
equalTo(new BytesRef("{\"@timestamp\":\"2012-02-13\",\"field\":\"value1\"}"))
735-
);
752+
if (IndexSettings.RECOVERY_USE_SYNTHETIC_SOURCE.isEnabled() == false) {
753+
// TODO: remove this if branch when removing the 'index_recovery_use_synthetic_source' feature flag
754+
assertNotNull(doc.rootDoc().getField("_recovery_source"));
755+
assertThat(
756+
doc.rootDoc().getField("_recovery_source").binaryValue(),
757+
equalTo(new BytesRef("{\"@timestamp\":\"2012-02-13\",\"field\":\"value1\"}"))
758+
);
759+
} else {
760+
assertNull(doc.rootDoc().getField("_recovery_source"));
761+
}
736762
}
737763
{
738764
Settings settings = Settings.builder()
@@ -776,11 +802,16 @@ public void testRecoverySourceWithTimeSeriesCustom() throws IOException {
776802
MapperService mapperService = createMapperService(settings, mappings);
777803
DocumentMapper docMapper = mapperService.documentMapper();
778804
ParsedDocument doc = docMapper.parse(source("123", b -> b.field("@timestamp", "2012-02-13").field("field", "value1"), null));
779-
assertNotNull(doc.rootDoc().getField("_recovery_source"));
780-
assertThat(
781-
doc.rootDoc().getField("_recovery_source").binaryValue(),
782-
equalTo(new BytesRef("{\"@timestamp\":\"2012-02-13\",\"field\":\"value1\"}"))
783-
);
805+
if (IndexSettings.RECOVERY_USE_SYNTHETIC_SOURCE.isEnabled() == false) {
806+
// TODO: remove this if branch when removing the 'index_recovery_use_synthetic_source' feature flag
807+
assertNotNull(doc.rootDoc().getField("_recovery_source"));
808+
assertThat(
809+
doc.rootDoc().getField("_recovery_source").binaryValue(),
810+
equalTo(new BytesRef("{\"@timestamp\":\"2012-02-13\",\"field\":\"value1\"}"))
811+
);
812+
} else {
813+
assertNull(doc.rootDoc().getField("_recovery_source"));
814+
}
784815
}
785816
{
786817
Settings settings = Settings.builder()

‎test/framework/src/main/java/org/elasticsearch/index/mapper/MapperServiceTestCase.java‎

Lines changed: 5 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -884,8 +884,11 @@ protected void validateRoundTripReader(String syntheticSource, DirectoryReader r
884884
throws IOException {
885885
assertReaderEquals(
886886
"round trip " + syntheticSource,
887-
new FieldMaskingReader(SourceFieldMapper.RECOVERY_SOURCE_NAME, reader),
888-
new FieldMaskingReader(SourceFieldMapper.RECOVERY_SOURCE_NAME, roundTripReader)
887+
new FieldMaskingReader(Set.of(SourceFieldMapper.RECOVERY_SOURCE_NAME, SourceFieldMapper.RECOVERY_SOURCE_SIZE_NAME), reader),
888+
new FieldMaskingReader(
889+
Set.of(SourceFieldMapper.RECOVERY_SOURCE_NAME, SourceFieldMapper.RECOVERY_SOURCE_SIZE_NAME),
890+
roundTripReader
891+
)
889892
);
890893
}
891894

‎x-pack/plugin/logsdb/build.gradle‎

Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -44,3 +44,13 @@ tasks.named("javaRestTest").configure {
4444
tasks.named('yamlRestTest') {
4545
usesDefaultDistribution()
4646
}
47+
48+
tasks.named("yamlRestTest") {
49+
if (buildParams.isSnapshotBuild() == false) {
50+
systemProperty 'tests.rest.blacklist', [
51+
"60_synthetic_source_recovery/*"
52+
].join(',')
53+
}
54+
}
55+
56+

0 commit comments

Comments
 (0)