Fix potential concurrent modification exception in DiscoveryNodeFilters

CHANGELOG.md

@@ -33,6 +33,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 - Fix pull-based ingestion out-of-bounds offset scenarios and remove persisted offsets ([#19607](https://github.com/opensearch-project/OpenSearch/pull/19607))
 - Fix issue with updating core with a patch number other than 0 ([#19377](https://github.com/opensearch-project/OpenSearch/pull/19377))
 - [Java Agent] Allow JRT protocol URLs in protection domain extraction ([#19683](https://github.com/opensearch-project/OpenSearch/pull/19683))
+- Fix potential concurrent modification exception when updating allocation filters ([#19701])(https://github.com/opensearch-project/OpenSearch/pull/19701))
 
 ### Dependencies
 - Update to Gradle 9.1 ([#19575](https://github.com/opensearch-project/OpenSearch/pull/19575))

server/src/main/java/org/opensearch/cluster/node/DiscoveryNodeFilters.java

@@ -97,7 +97,7 @@ public static DiscoveryNodeFilters buildOrUpdateFromKeyValue(
             updated = new DiscoveryNodeFilters(opType, new HashMap<>());
         } else {
             assert opType == original.opType : "operation type should match with node filter parameter";
-            updated = new DiscoveryNodeFilters(original.opType, original.filters);
+            updated = new DiscoveryNodeFilters(original.opType, new HashMap<>(original.filters));
         }
         for (Map.Entry<String, String> entry : filters.entrySet()) {
             String[] values = Strings.tokenizeToStringArray(entry.getValue(), ",");

server/src/main/java/org/opensearch/cluster/routing/allocation/decider/FilterAllocationDecider.java

@@ -193,7 +193,7 @@ public Decision shouldAutoExpandToNode(IndexMetadata indexMetadata, DiscoveryNod
     @Override
     public Decision canAllocateAnyShardToNode(RoutingNode node, RoutingAllocation allocation) {
         Decision decision = shouldClusterFilter(node.node(), allocation);
-        return decision != null && decision == Decision.NO ? decision : Decision.ALWAYS;
+        return decision == Decision.NO ? decision : Decision.ALWAYS;
     }
 
     private Decision shouldFilter(ShardRouting shardRouting, DiscoveryNode node, RoutingAllocation allocation) {
@@ -258,6 +258,13 @@ private Decision shouldIndexFilter(IndexMetadata indexMd, DiscoveryNode node, Ro
     }
 
     private Decision shouldClusterFilter(DiscoveryNode node, RoutingAllocation allocation) {
+        // Copy values to local variables so we're not null-checking on volatile fields.
+        // The value of a volatile field could change from non-null to null between the
+        // check and its usage.
+        DiscoveryNodeFilters clusterRequireFilters = this.clusterRequireFilters;
+        DiscoveryNodeFilters clusterIncludeFilters = this.clusterIncludeFilters;
+        DiscoveryNodeFilters clusterExcludeFilters = this.clusterExcludeFilters;
+
         if (clusterRequireFilters != null) {
             if (clusterRequireFilters.match(node) == false) {
                 return allocation.decision(

server/src/test/java/org/opensearch/cluster/node/DiscoveryNodeFiltersTests.java

@@ -47,6 +47,12 @@
 import java.util.HashMap;
 import java.util.List;
 import java.util.Map;
+import java.util.concurrent.ExecutionException;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Executors;
+import java.util.concurrent.Future;
+import java.util.concurrent.atomic.AtomicBoolean;
+import java.util.concurrent.atomic.AtomicReference;
 
 import static java.util.Collections.emptyMap;
 import static java.util.Collections.emptySet;
@@ -354,6 +360,67 @@ public void testOpTypeMismatch() {
         }
     }
 
+    public void testConcurrentModification() {
+        AtomicReference<DiscoveryNodeFilters> filters = new AtomicReference<>();
+        AtomicBoolean keepRunning = new AtomicBoolean(true);
+        int count = 200; // I can pretty reliably reproduce failures with 200 iterations, but not 150
+        try (ExecutorService executor = Executors.newFixedThreadPool(2)) {
+            // Thread 1: repeatedly update the filters
+            Future<?> t1 = executor.submit(() -> {
+                try {
+                    for (int i = 0; i < count && keepRunning.get(); i++) {
+                        Settings.Builder settingsBuilder = Settings.builder().put("xxx._id", "id" + i);
+                        if (i % 2 == 0) {
+                            settingsBuilder.put("xxx._name", "");
+                        } else {
+                            settingsBuilder.put("xxx._name", "name" + i);
+                        }
+                        DiscoveryNodeFilters newFilters = buildOrUpdateFromSettings(filters.get(), OR, "xxx.", settingsBuilder.build());
+                        filters.set(newFilters);
+                        try {
+                            Thread.sleep(1);
+                        } catch (InterruptedException e) {
+                            throw new RuntimeException(e);
+                        }
+                    }
+                } finally {
+                    keepRunning.set(false);
+                }
+            });
+
+            // Thread 2: repeatedly read and match nodes against the filters
+            Future<?> t2 = executor.submit(() -> {
+                try {
+                    for (int i = 0; i < count && keepRunning.get(); i++) {
+                        DiscoveryNodeFilters currentFilters = filters.get();
+                        if (currentFilters != null) {
+                            DiscoveryNode node = new DiscoveryNode(
+                                "name" + i,
+                                "id" + i,
+                                buildNewFakeTransportAddress(),
+                                emptyMap(),
+                                emptySet(),
+                                Version.CURRENT
+                            );
+                            currentFilters.match(node);
+                        }
+                        try {
+                            Thread.sleep(1);
+                        } catch (InterruptedException e) {
+                            throw new RuntimeException(e);
+                        }
+                    }
+                } finally {
+                    keepRunning.set(false);
+                }
+            });
+            t2.get();
+            t1.get();
+        } catch (ExecutionException | InterruptedException e) {
+            throw new RuntimeException(e);
+        }
+    }
+
     private Settings shuffleSettings(Settings source) {
         Settings.Builder settings = Settings.builder();
         List<String> keys = new ArrayList<>(source.keySet());