Merge remote-tracking branch 'upstream/main' into tiramisu-stats-api-final

Signed-off-by: Peter Alfonsi <[email protected]>

Author: Peter Alfonsi

.github/workflows/assemble.yml

@@ -8,7 +8,7 @@ jobs:
     strategy:
       matrix:
         java: [ 11, 17, 21 ]
-        os: [ubuntu-latest, windows-latest, macos-latest]
+        os: [ubuntu-latest, windows-latest, macos-13]
     steps:
       - uses: actions/checkout@v4
       - name: Set up JDK ${{ matrix.java }}
@@ -18,10 +18,7 @@ jobs:
           distribution: temurin
       - name: Setup docker (missing on MacOS)
         if: runner.os == 'macos'
-        run: |
-          brew install docker
-          colima start
-          sudo ln -sf $HOME/.colima/default/docker.sock /var/run/docker.sock
+        uses: douglascamata/setup-docker-macos-action@main
       - name: Run Gradle (assemble)
         run: |
           ./gradlew assemble --parallel --no-build-cache -PDISABLE_BUILD_CACHE

.github/workflows/precommit.yml

@@ -8,7 +8,7 @@ jobs:
     strategy:
       matrix:
         java: [ 11, 17, 21 ]
-        os: [ubuntu-latest, windows-latest, macos-latest]
+        os: [ubuntu-latest, windows-latest, macos-13]
     steps:
       - uses: actions/checkout@v4
       - name: Set up JDK ${{ matrix.java }}

CHANGELOG.md

@@ -25,6 +25,11 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - [Tiered Caching] Expose new cache stats API ([#13237](https://github.com/opensearch-project/OpenSearch/pull/13237))
 - [Streaming Indexing] Ensure support of the new transport by security plugin ([#13174](https://github.com/opensearch-project/OpenSearch/pull/13174))
 - Add cluster setting to dynamically configure the buckets for filter rewrite optimization. ([#13179](https://github.com/opensearch-project/OpenSearch/pull/13179))
+- [Tiered Caching] Gate new stats logic behind FeatureFlags.PLUGGABLE_CACHE ([#13238](https://github.com/opensearch-project/OpenSearch/pull/13238))
+- [Tiered Caching] Add a dynamic setting to disable/enable disk cache. ([#13373](https://github.com/opensearch-project/OpenSearch/pull/13373))
+- [Remote Store] Add capability of doing refresh as determined by the translog ([#12992](https://github.com/opensearch-project/OpenSearch/pull/12992))
+- [Tiered caching] Make Indices Request Cache Stale Key Mgmt Threshold setting dynamic ([#12941](https://github.com/opensearch-project/OpenSearch/pull/12941))
+- Batch mode for async fetching shard information in GatewayAllocator for unassigned shards ([#8746](https://github.com/opensearch-project/OpenSearch/pull/8746))
 
 ### Dependencies
 - Bump `org.apache.commons:commons-configuration2` from 2.10.0 to 2.10.1 ([#12896](https://github.com/opensearch-project/OpenSearch/pull/12896))
@@ -46,6 +51,8 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - Bump `com.github.spullara.mustache.java:compiler` from 0.9.10 to 0.9.11 ([#13329](https://github.com/opensearch-project/OpenSearch/pull/13329))
 - Bump `jakarta.enterprise:jakarta.enterprise.cdi-api` from 4.0.1 to 4.1.0 ([#13328](https://github.com/opensearch-project/OpenSearch/pull/13328))
 - Bump `com.google.api.grpc:proto-google-iam-v1` from 0.12.0 to 1.33.0 ([#13332](https://github.com/opensearch-project/OpenSearch/pull/13332))
+- OpenJDK Update (April 2024 Patch releases), update to Eclipse Temurin 21.0.3+9 ([#13389](https://github.com/opensearch-project/OpenSearch/pull/13389))
+- Bump `com.squareup.okio:okio` from 3.8.0 to 3.9.0 ([#12997](https://github.com/opensearch-project/OpenSearch/pull/12997))
 
 ### Changed
 - [BWC and API enforcement] Enforcing the presence of API annotations at build time ([#12872](https://github.com/opensearch-project/OpenSearch/pull/12872))
@@ -54,6 +61,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - Ignoring unavailable shards during search request execution with ignore_available parameter ([#13298](https://github.com/opensearch-project/OpenSearch/pull/13298))
 - Refactoring globMatch using simpleMatchWithNormalizedStrings from Regex ([#13104](https://github.com/opensearch-project/OpenSearch/pull/13104))
 - [BWC and API enforcement] Reconsider the breaking changes check policy to detect breaking changes against released versions ([#13292](https://github.com/opensearch-project/OpenSearch/pull/13292))
+- Switch to macos-13 runner for precommit and assemble github actions due to macos-latest is now arm64 ([#13412](https://github.com/opensearch-project/OpenSearch/pull/13412))
 
 ### Deprecated
 

buildSrc/src/main/java/org/opensearch/gradle/test/DistroTestPlugin.java

@@ -77,9 +77,9 @@
 import java.util.stream.Stream;
 
 public class DistroTestPlugin implements Plugin<Project> {
-    private static final String SYSTEM_JDK_VERSION = "21.0.2+13";
+    private static final String SYSTEM_JDK_VERSION = "21.0.3+9";
     private static final String SYSTEM_JDK_VENDOR = "adoptium";
-    private static final String GRADLE_JDK_VERSION = "21.0.2+13";
+    private static final String GRADLE_JDK_VERSION = "21.0.3+9";
     private static final String GRADLE_JDK_VENDOR = "adoptium";
 
     // all distributions used by distro tests. this is temporary until tests are per distribution

buildSrc/version.properties

@@ -2,7 +2,7 @@ opensearch        = 3.0.0
 lucene            = 9.11.0-snapshot-fb97840
 
 bundled_jdk_vendor = adoptium
-bundled_jdk = 21.0.2+13
+bundled_jdk = 21.0.3+9
 
 # optional dependencies
 spatial4j         = 0.7

modules/cache-common/src/internalClusterTest/java/org.opensearch.cache.common.tier/TieredSpilloverCacheIT.java

@@ -425,6 +425,121 @@ public void testWithExplicitCacheClear() throws Exception {
         }, 1, TimeUnit.SECONDS);
     }
 
+    public void testWithDynamicDiskCacheSetting() throws Exception {
+        int onHeapCacheSizeInBytes = 10; // Keep it low so that all items are cached onto disk.
+        internalCluster().startNode(
+            Settings.builder()
+                .put(defaultSettings(onHeapCacheSizeInBytes + "b"))
+                .put(INDICES_CACHE_CLEAN_INTERVAL_SETTING.getKey(), new TimeValue(1))
+                .build()
+        );
+        Client client = client();
+        assertAcked(
+            client.admin()
+                .indices()
+                .prepareCreate("index")
+                .setMapping("k", "type=keyword")
+                .setSettings(
+                    Settings.builder()
+                        .put(IndicesRequestCache.INDEX_CACHE_REQUEST_ENABLED_SETTING.getKey(), true)
+                        .put(IndexMetadata.SETTING_NUMBER_OF_SHARDS, 1)
+                        .put(IndexMetadata.SETTING_NUMBER_OF_REPLICAS, 0)
+                        .put("index.refresh_interval", -1)
+                )
+                .get()
+        );
+        // Update took time policy to zero so that all entries are eligible to be cached on disk.
+        ClusterUpdateSettingsRequest updateSettingsRequest = new ClusterUpdateSettingsRequest().transientSettings(
+            Settings.builder()
+                .put(
+                    TieredSpilloverCacheSettings.TOOK_TIME_POLICY_CONCRETE_SETTINGS_MAP.get(CacheType.INDICES_REQUEST_CACHE).getKey(),
+                    new TimeValue(0, TimeUnit.MILLISECONDS)
+                )
+                .build()
+        );
+        assertAcked(internalCluster().client().admin().cluster().updateSettings(updateSettingsRequest).get());
+        int numberOfIndexedItems = randomIntBetween(5, 10);
+        for (int iterator = 0; iterator < numberOfIndexedItems; iterator++) {
+            indexRandom(true, client.prepareIndex("index").setSource("k" + iterator, "hello" + iterator));
+        }
+        ensureSearchable("index");
+        refreshAndWaitForReplication();
+        // Force merge the index to ensure there can be no background merges during the subsequent searches that would invalidate the cache
+        ForceMergeResponse forceMergeResponse = client.admin().indices().prepareForceMerge("index").setFlush(true).get();
+        OpenSearchAssertions.assertAllSuccessful(forceMergeResponse);
+        long perQuerySizeInCacheInBytes = -1;
+        // Step 1: Hit some queries. We will see misses and queries will be cached(onto disk cache) for subsequent
+        // requests.
+        for (int iterator = 0; iterator < numberOfIndexedItems; iterator++) {
+            SearchResponse resp = client.prepareSearch("index")
+                .setRequestCache(true)
+                .setQuery(QueryBuilders.termQuery("k" + iterator, "hello" + iterator))
+                .get();
+            if (perQuerySizeInCacheInBytes == -1) {
+                RequestCacheStats requestCacheStats = getRequestCacheStats(client, "index");
+                perQuerySizeInCacheInBytes = requestCacheStats.getMemorySizeInBytes();
+            }
+            assertSearchResponse(resp);
+        }
+
+        RequestCacheStats requestCacheStats = getRequestCacheStats(client, "index");
+        assertEquals(numberOfIndexedItems * perQuerySizeInCacheInBytes, requestCacheStats.getMemorySizeInBytes());
+        assertEquals(numberOfIndexedItems, requestCacheStats.getMissCount());
+        assertEquals(0, requestCacheStats.getHitCount());
+        assertEquals(0, requestCacheStats.getEvictions());
+
+        // Step 2: Hit same queries again. We will see hits now.
+        for (int iterator = 0; iterator < numberOfIndexedItems; iterator++) {
+            SearchResponse resp = client.prepareSearch("index")
+                .setRequestCache(true)
+                .setQuery(QueryBuilders.termQuery("k" + iterator, "hello" + iterator))
+                .get();
+            assertSearchResponse(resp);
+        }
+        requestCacheStats = getRequestCacheStats(client, "index");
+        assertEquals(numberOfIndexedItems * perQuerySizeInCacheInBytes, requestCacheStats.getMemorySizeInBytes());
+        assertEquals(numberOfIndexedItems, requestCacheStats.getMissCount());
+        assertEquals(numberOfIndexedItems, requestCacheStats.getHitCount());
+        assertEquals(0, requestCacheStats.getEvictions());
+        long lastKnownHitCount = requestCacheStats.getHitCount();
+        long lastKnownMissCount = requestCacheStats.getMissCount();
+
+        // Step 3: Turn off disk cache now. And hit same queries again. We should not see hits now as all queries
+        // were cached onto disk cache.
+        updateSettingsRequest = new ClusterUpdateSettingsRequest().transientSettings(
+            Settings.builder()
+                .put(TieredSpilloverCacheSettings.DISK_CACHE_ENABLED_SETTING_MAP.get(CacheType.INDICES_REQUEST_CACHE).getKey(), false)
+                .build()
+        );
+        assertAcked(internalCluster().client().admin().cluster().updateSettings(updateSettingsRequest).get());
+
+        for (int iterator = 0; iterator < numberOfIndexedItems; iterator++) {
+            SearchResponse resp = client.prepareSearch("index")
+                .setRequestCache(true)
+                .setQuery(QueryBuilders.termQuery("k" + iterator, "hello" + iterator))
+                .get();
+            assertSearchResponse(resp);
+        }
+        requestCacheStats = getRequestCacheStats(client, "index");
+        assertEquals(numberOfIndexedItems * perQuerySizeInCacheInBytes, requestCacheStats.getMemorySizeInBytes()); //
+        // Still shows disk cache entries as explicit clear or invalidation is required to clean up disk cache.
+        assertEquals(lastKnownMissCount + numberOfIndexedItems, requestCacheStats.getMissCount());
+        assertEquals(0, lastKnownHitCount - requestCacheStats.getHitCount()); // No new hits being seen.
+        lastKnownMissCount = requestCacheStats.getMissCount();
+        lastKnownHitCount = requestCacheStats.getHitCount();
+
+        // Step 4: Invalidate entries via refresh.
+        // Explicit refresh would invalidate cache entries.
+        refreshAndWaitForReplication();
+        assertBusy(() -> {
+            // Explicit refresh should clear up cache entries
+            assertTrue(getRequestCacheStats(client, "index").getMemorySizeInBytes() == 0);
+        }, 1, TimeUnit.SECONDS);
+        requestCacheStats = getRequestCacheStats(client, "index");
+        assertEquals(0, lastKnownMissCount - requestCacheStats.getMissCount());
+        assertEquals(0, lastKnownHitCount - requestCacheStats.getHitCount());
+    }
+
     private RequestCacheStats getRequestCacheStats(Client client, String indexName) {
         return client.admin().indices().prepareStats(indexName).setRequestCache(true).get().getTotal().getRequestCache();
     }