Merge branch 'startree-jun27' into star_tree_codec

.github/workflows/gradle-check.yml

@@ -113,6 +113,7 @@ jobs:
         if: success()
         uses: codecov/codecov-action@v4
         with:
+          token: ${{ secrets.CODECOV_TOKEN }}
           files: ./codeCoverage.xml
 
       - name: Create Comment Success

CHANGELOG.md

@@ -9,6 +9,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - [Remote Store] Rate limiter for remote store low priority uploads ([#14374](https://github.com/opensearch-project/OpenSearch/pull/14374/))
 - Apply the date histogram rewrite optimization to range aggregation ([#13865](https://github.com/opensearch-project/OpenSearch/pull/13865))
 - [Writable Warm] Add composite directory implementation and integrate it with FileCache ([12782](https://github.com/opensearch-project/OpenSearch/pull/12782))
+- Fix race condition while parsing derived fields from search definition ([14445](https://github.com/opensearch-project/OpenSearch/pull/14445))
 
 ### Dependencies
 - Bump `org.gradle.test-retry` from 1.5.8 to 1.5.9 ([#13442](https://github.com/opensearch-project/OpenSearch/pull/13442))
@@ -25,6 +26,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - Bump `azure-identity` from 1.11.4 to 1.13.0, Bump `msal4j` from 1.14.3 to 1.15.1, Bump `msal4j-persistence-extension` from 1.2.0 to 1.3.0 ([#14506](https://github.com/opensearch-project/OpenSearch/pull/14506))
 
 ### Changed
+- [Tiered Caching] Move query recomputation logic outside write lock ([#14187](https://github.com/opensearch-project/OpenSearch/pull/14187))
 - unsignedLongRangeQuery now returns MatchNoDocsQuery if the lower bounds are greater than the upper bounds ([#14416](https://github.com/opensearch-project/OpenSearch/pull/14416))
 - Updated the `indices.query.bool.max_clause_count` setting from being static to dynamically updateable ([#13568](https://github.com/opensearch-project/OpenSearch/pull/13568))
 - Make the class CommunityIdProcessor final ([#14448](https://github.com/opensearch-project/OpenSearch/pull/14448))
@@ -41,7 +43,9 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - Fixed rest-high-level client searchTemplate & mtermVectors endpoints to have a leading slash ([#14465](https://github.com/opensearch-project/OpenSearch/pull/14465))
 - Write shard level metadata blob when snapshotting searchable snapshot indexes ([#13190](https://github.com/opensearch-project/OpenSearch/pull/13190))
 - Fix aggs result of NestedAggregator with sub NestedAggregator ([#13324](https://github.com/opensearch-project/OpenSearch/pull/13324))
+- Fix fs info reporting negative available size ([#11573](https://github.com/opensearch-project/OpenSearch/pull/11573))
 - Add ListPitInfo::getKeepAlive() getter ([#14495](https://github.com/opensearch-project/OpenSearch/pull/14495))
+- Fix FuzzyQuery in keyword field will use IndexOrDocValuesQuery when both of index and doc_value are true ([#14378](https://github.com/opensearch-project/OpenSearch/pull/14378))
 
 ### Security
 

TESTING.md

@@ -17,6 +17,8 @@ OpenSearch uses [jUnit](https://junit.org/junit5/) for testing, it also uses ran
   - [Miscellaneous](#miscellaneous)
 - [Running verification tasks](#running-verification-tasks)
 - [Testing the REST layer](#testing-the-rest-layer)
+  - [Running REST Tests Against An External Cluster](#running-rest-tests-against-an-external-cluster)
+  - [Debugging REST Tests](#debugging-rest-tests)
 - [Testing packaging](#testing-packaging)
   - [Testing packaging on Windows](#testing-packaging-on-windows)
   - [Testing VMs are disposable](#testing-vms-are-disposable)
@@ -272,7 +274,18 @@ yamlRestTest’s and javaRestTest’s are easy to identify, since they are found
 
 If in doubt about which command to use, simply run <gradle path>:check
 
-Note that the REST tests, like all the integration tests, can be run against an external cluster by specifying the `tests.cluster` property, which if present needs to contain a comma separated list of nodes to connect to (e.g. localhost:9300).
+## Running REST Tests Against An External Cluster
+
+Note that the REST tests, like all the integration tests, can be run against an external cluster by specifying the following properties `tests.cluster`, `tests.rest.cluster`, `tests.clustername`. Use a comma separated list of node properties for the multi-node cluster.
+
+For example :
+
+    ./gradlew :rest-api-spec:yamlRestTest \
+      -Dtests.cluster=localhost:9200 -Dtests.rest.cluster=localhost:9200 -Dtests.clustername=opensearch
+
+## Debugging REST Tests
+
+You can launch a local OpenSearch cluster in debug mode following [Launching and debugging from an IDE](#launching-and-debugging-from-an-ide), and run your REST tests against that following [Running REST Tests Against An External Cluster](#running-rest-tests-against-an-external-cluster).
 
 # Testing packaging
 

TRIAGING.md

@@ -14,7 +14,7 @@ Each meeting we seek to address all new issues. However, should we run out of ti
 
 ### How do I join a Triage meeting?
 
- Check the [OpenSearch Meetup Group](https://www.meetup.com/opensearch/) for the latest schedule and details for joining each meeting. Each component area has its own meetup series: [Search](https://www.meetup.com/opensearch/events/300929493/), [Storage](https://www.meetup.com/opensearch/events/299907409/), [Cluster Manager](https://www.meetup.com/opensearch/events/301082218/), and [Core](https://www.meetup.com/opensearch/events/301061009/).
+ Check the [OpenSearch Meetup Group](https://www.meetup.com/opensearch/) for the latest schedule and details for joining each meeting. Each component area has its own meetup series: [Search](https://www.meetup.com/opensearch/events/300929493/), [Storage](https://www.meetup.com/opensearch/events/299907409/), [Cluster Manager](https://www.meetup.com/opensearch/events/301082218/), [Indexing](https://www.meetup.com/opensearch/events/301734024/), and [Core](https://www.meetup.com/opensearch/events/301061009/).
 
 After joining the virtual meeting, you can enable your video / voice to join the discussion.  If you do not have a webcam or microphone available, you can still join in via the text chat.
 
@@ -29,9 +29,10 @@ Meeting structure may vary slightly, but the general structure is as follows:
 3. **Announcements:** Any announcements will be made at the beginning of the meeting.
 4. **Review of New Issues:** We start by reviewing all untriaged issues. Each meeting has a label-based search to find relevant issues:
    - [Search](https://github.com/opensearch-project/OpenSearch/issues?q=is%3Aissue+is%3Aopen+label%3Auntriaged+label%3A%22Search%22%2C%22Search%3ARemote+Search%22%2C%22Search%3AResiliency%22%2C%22Search%3APerformance%22%2C%22Search%3ARelevance%22%2C%22Search%3AAggregations%22%2C%22Search%3AQuery+Capabilities%22%2C%22Search%3AQuery+Insights%22%2C%22Search%3ASearchable+Snapshots%22%2C%22Search%3AUser+Behavior+Insights%22)
+   - [Indexing](https://github.com/opensearch-project/OpenSearch/issues?q=is%3Aissue+is%3Aopen+label%3Auntriaged+label%3A%22Indexing%3AReplication%22%2C%22Indexing%22%2C%22Indexing%3APerformance%22%2C%22Indexing+%26+Search%22%2C)
    - [Storage](https://github.com/opensearch-project/OpenSearch/issues?q=is%3Aissue+is%3Aopen+label%3Auntriaged+label%3AStorage%2C%22Storage%3AResiliency%22%2C%22Storage%3APerformance%22%2C%22Storage%3ASnapshots%22%2C%22Storage%3ARemote%22%2C%22Storage%3ADurability%22)
    - [Cluster Manager](https://github.com/opensearch-project/OpenSearch/issues?q=is%3Aissue+is%3Aopen+label%3Auntriaged+label%3A%22Cluster+Manager%22%2C%22ClusterManager%3ARemoteState%22)
-   - [Core](https://github.com/opensearch-project/OpenSearch/issues?q=is%3Aissue+is%3Aopen+label%3Auntriaged+-label%3A%22Search%22%2C%22Search%3ARemote+Search%22%2C%22Search%3AResiliency%22%2C%22Search%3APerformance%22%2C%22Search%3ARelevance%22%2C%22Search%3AAggregations%22%2C%22Search%3AQuery+Capabilities%22%2C%22Search%3AQuery+Insights%22%2C%22Search%3ASearchable+Snapshots%22%2C%22Search%3AUser+Behavior+Insights%22%2C%22Storage%22%2C%22Storage%3AResiliency%22%2C%22Storage%3APerformance%22%2C%22Storage%3ASnapshots%22%2C%22Storage%3ARemote%22%2C%22Storage%3ADurability%22%2C%22Cluster+Manager%22%2C%22ClusterManager%3ARemoteState%22)
+   - [Core](https://github.com/opensearch-project/OpenSearch/issues?q=is%3Aissue+is%3Aopen+label%3Auntriaged+-label%3A%22Search%22%2C%22Search%3ARemote+Search%22%2C%22Search%3AResiliency%22%2C%22Search%3APerformance%22%2C%22Search%3ARelevance%22%2C%22Search%3AAggregations%22%2C%22Search%3AQuery+Capabilities%22%2C%22Search%3AQuery+Insights%22%2C%22Search%3ASearchable+Snapshots%22%2C%22Search%3AUser+Behavior+Insights%22%2C%22Storage%22%2C%22Storage%3AResiliency%22%2C%22Storage%3APerformance%22%2C%22Storage%3ASnapshots%22%2C%22Storage%3ARemote%22%2C%22Storage%3ADurability%22%2C%22Cluster+Manager%22%2C%22ClusterManager%3ARemoteState%22%2C%22Indexing%3AReplication%22%2C%22Indexing%22%2C%22Indexing%3APerformance%22%2C%22Indexing+%26+Search%22)
 5. **Attendee Requests:** An opportunity for any meeting member to request consideration of an issue or pull request.
 6. **Open Discussion:** Attendees can bring up any topics not already covered by filed issues or pull requests.
 7. **Review of Old Untriaged Issues:** Time permitting, each meeting will look at all [untriaged issues older than 14 days](https://github.com/opensearch-project/OpenSearch/issues?q=is%3Aissue+is%3Aopen+label%3Auntriaged+created%3A%3C2024-05-20) to prevent issues from falling through the cracks (note the GitHub API does not allow for relative times, so the date in this search must be updated every meeting).

distribution/src/config/opensearch.yml

@@ -125,3 +125,7 @@ ${path.logs}
 # Gates the functionality of enabling Opensearch to use pluggable caches with respective store names via setting.
 #
 #opensearch.experimental.feature.pluggable.caching.enabled: false
+#
+# Gates the functionality of star tree index, which improves the performance of search aggregations.
+#
+#opensearch.experimental.feature.composite_index.star_tree.enabled: true

modules/cache-common/src/main/java/org/opensearch/cache/common/tier/TieredSpilloverCache.java

@@ -8,6 +8,8 @@
 
 package org.opensearch.cache.common.tier;
 
+import org.apache.logging.log4j.LogManager;
+import org.apache.logging.log4j.Logger;
 import org.opensearch.cache.common.policy.TookTimePolicy;
 import org.opensearch.common.annotation.ExperimentalApi;
 import org.opensearch.common.cache.CacheType;
@@ -35,9 +37,13 @@
 import java.util.Map;
 import java.util.NoSuchElementException;
 import java.util.Objects;
+import java.util.concurrent.CompletableFuture;
+import java.util.concurrent.ConcurrentHashMap;
+import java.util.concurrent.ExecutionException;
 import java.util.concurrent.atomic.AtomicBoolean;
 import java.util.concurrent.locks.ReadWriteLock;
 import java.util.concurrent.locks.ReentrantReadWriteLock;
+import java.util.function.BiFunction;
 import java.util.function.Function;
 import java.util.function.Predicate;
 import java.util.function.ToLongBiFunction;
@@ -61,6 +67,7 @@ public class TieredSpilloverCache<K, V> implements ICache<K, V> {
 
     // Used to avoid caching stale entries in lower tiers.
     private static final List<RemovalReason> SPILLOVER_REMOVAL_REASONS = List.of(RemovalReason.EVICTED, RemovalReason.CAPACITY);
+    private static final Logger logger = LogManager.getLogger(TieredSpilloverCache.class);
 
     private final ICache<K, V> diskCache;
     private final ICache<K, V> onHeapCache;
@@ -86,6 +93,12 @@ public class TieredSpilloverCache<K, V> implements ICache<K, V> {
     private final Map<ICache<K, V>, TierInfo> caches;
     private final List<Predicate<V>> policies;
 
+    /**
+     * This map is used to handle concurrent requests for same key in computeIfAbsent() to ensure we load the value
+     * only once.
+     */
+    Map<ICacheKey<K>, CompletableFuture<Tuple<ICacheKey<K>, V>>> completableFutureMap = new ConcurrentHashMap<>();
+
     TieredSpilloverCache(Builder<K, V> builder) {
         Objects.requireNonNull(builder.onHeapCacheFactory, "onHeap cache builder can't be null");
         Objects.requireNonNull(builder.diskCacheFactory, "disk cache builder can't be null");
@@ -190,10 +203,7 @@ public V computeIfAbsent(ICacheKey<K> key, LoadAwareCacheLoader<ICacheKey<K>, V>
             // Add the value to the onHeap cache. We are calling computeIfAbsent which does another get inside.
             // This is needed as there can be many requests for the same key at the same time and we only want to load
             // the value once.
-            V value = null;
-            try (ReleasableLock ignore = writeLock.acquire()) {
-                value = onHeapCache.computeIfAbsent(key, loader);
-            }
+            V value = compute(key, loader);
             // Handle stats
             if (loader.isLoaded()) {
                 // The value was just computed and added to the cache by this thread. Register a miss for the heap cache, and the disk cache
@@ -222,6 +232,57 @@ public V computeIfAbsent(ICacheKey<K> key, LoadAwareCacheLoader<ICacheKey<K>, V>
         return cacheValueTuple.v1();
     }
 
+    private V compute(ICacheKey<K> key, LoadAwareCacheLoader<ICacheKey<K>, V> loader) throws Exception {
+        // Only one of the threads will succeed putting a future into map for the same key.
+        // Rest will fetch existing future and wait on that to complete.
+        CompletableFuture<Tuple<ICacheKey<K>, V>> future = completableFutureMap.putIfAbsent(key, new CompletableFuture<>());
+        // Handler to handle results post processing. Takes a tuple<key, value> or exception as an input and returns
+        // the value. Also before returning value, puts the value in cache.
+        BiFunction<Tuple<ICacheKey<K>, V>, Throwable, Void> handler = (pair, ex) -> {
+            if (pair != null) {
+                try (ReleasableLock ignore = writeLock.acquire()) {
+                    onHeapCache.put(pair.v1(), pair.v2());
+                } catch (Exception e) {
+                    // TODO: Catch specific exceptions to know whether this resulted from cache or underlying removal
+                    // listeners/stats. Needs better exception handling at underlying layers.For now swallowing
+                    // exception.
+                    logger.warn("Exception occurred while putting item onto heap cache", e);
+                }
+            } else {
+                if (ex != null) {
+                    logger.warn("Exception occurred while trying to compute the value", ex);
+                }
+            }
+            completableFutureMap.remove(key); // Remove key from map as not needed anymore.
+            return null;
+        };
+        V value = null;
+        if (future == null) {
+            future = completableFutureMap.get(key);
+            future.handle(handler);
+            try {
+                value = loader.load(key);
+            } catch (Exception ex) {
+                future.completeExceptionally(ex);
+                throw new ExecutionException(ex);
+            }
+            if (value == null) {
+                NullPointerException npe = new NullPointerException("Loader returned a null value");
+                future.completeExceptionally(npe);
+                throw new ExecutionException(npe);
+            } else {
+                future.complete(new Tuple<>(key, value));
+            }
+        } else {
+            try {
+                value = future.get().v2();
+            } catch (InterruptedException ex) {
+                throw new IllegalStateException(ex);
+            }
+        }
+        return value;
+    }
+
     @Override
     public void invalidate(ICacheKey<K> key) {
         // We are trying to invalidate the key from all caches though it would be present in only of them.
@@ -328,12 +389,22 @@ void handleRemovalFromHeapTier(RemovalNotification<ICacheKey<K>, V> notification
         ICacheKey<K> key = notification.getKey();
         boolean wasEvicted = SPILLOVER_REMOVAL_REASONS.contains(notification.getRemovalReason());
         boolean countEvictionTowardsTotal = false; // Don't count this eviction towards the cache's total if it ends up in the disk tier
-        if (caches.get(diskCache).isEnabled() && wasEvicted && evaluatePolicies(notification.getValue())) {
+        boolean exceptionOccurredOnDiskCachePut = false;
+        boolean canCacheOnDisk = caches.get(diskCache).isEnabled() && wasEvicted && evaluatePolicies(notification.getValue());
+        if (canCacheOnDisk) {
             try (ReleasableLock ignore = writeLock.acquire()) {
                 diskCache.put(key, notification.getValue()); // spill over to the disk tier and increment its stats
+            } catch (Exception ex) {
+                // TODO: Catch specific exceptions. Needs better exception handling. We are just swallowing exception
+                // in this case as it shouldn't cause upstream request to fail.
+                logger.warn("Exception occurred while putting item to disk cache", ex);
+                exceptionOccurredOnDiskCachePut = true;
             }
-            updateStatsOnPut(TIER_DIMENSION_VALUE_DISK, key, notification.getValue());
-        } else {
+            if (!exceptionOccurredOnDiskCachePut) {
+                updateStatsOnPut(TIER_DIMENSION_VALUE_DISK, key, notification.getValue());
+            }
+        }
+        if (!canCacheOnDisk || exceptionOccurredOnDiskCachePut) {
             // If the value is not going to the disk cache, send this notification to the TSC's removal listener
             // as the value is leaving the TSC entirely
             removalListener.onRemoval(notification);