This commit adds a new test CMS that always provides intra-merge parallelism (#13475)

@iverase has uncovered a potential issue with intra-merge CMS parallelism.

This commit helps expose this problem by forcing tests to use intra-merge parallelism instead of always (well, usually) delegating to a SameThreadExecutorService.

When intra-merge parallelism is used, norms, doc_values, stored_values, etc. are all merged in a separate thread than the thread that was used to construct their merge optimized instances.

This trips assertions numerous assertions in AssertingCodec.assertThread where we assume that the thread that called getMergeInstance() is also the thread getting the values to merge.

In addition to the better testing, this corrects poor merge state handling in the event of parallelism.

lucene/CHANGES.txt

@@ -333,6 +333,8 @@ New Features
 
 Improvements
 ---------------------
+* GITHUB#13475: Re-enable intra-merge parallelism except for terms, norms, and doc values.
+  Related to GITHUB#13478.  (Ben Trent)
 
 * GITHUB#13548: Refactor and javadoc update for KNN vector writer classes. (Patrick Zhai)
 

lucene/core/src/java/org/apache/lucene/index/MergeState.java

@@ -41,7 +41,7 @@ import org.apache.lucene.util.packed.PackedLongValues;
  *
  * @lucene.experimental
  */
-public class MergeState {
+public class MergeState implements Cloneable {
 
   /** Maps document IDs from old segments to document IDs in the new segment */
   public final DocMap[] docMaps;
@@ -302,4 +302,55 @@ public class MergeState {
     this.intraMergeTaskExecutor = intraMergeTaskExecutor;
     this.needsIndexSort = needsIndexSort;
   }
+
+  @Override
+  public MergeState clone() {
+    StoredFieldsReader[] storedFieldsReaders = this.storedFieldsReaders.clone();
+    TermVectorsReader[] termVectorsReaders = this.termVectorsReaders.clone();
+    NormsProducer[] normsProducers = this.normsProducers.clone();
+    DocValuesProducer[] docValuesProducers = this.docValuesProducers.clone();
+    FieldsProducer[] fieldsProducers = this.fieldsProducers.clone();
+    PointsReader[] pointsReaders = this.pointsReaders.clone();
+    KnnVectorsReader[] knnVectorsReaders = this.knnVectorsReaders.clone();
+    for (int i = 0; i < storedFieldsReaders.length; ++i) {
+      if (storedFieldsReaders[i] != null) {
+        storedFieldsReaders[i] = storedFieldsReaders[i].getMergeInstance();
+      }
+      if (termVectorsReaders[i] != null) {
+        termVectorsReaders[i] = termVectorsReaders[i].getMergeInstance();
+      }
+      if (normsProducers[i] != null) {
+        normsProducers[i] = normsProducers[i].getMergeInstance();
+      }
+      if (docValuesProducers[i] != null) {
+        docValuesProducers[i] = docValuesProducers[i].getMergeInstance();
+      }
+      if (fieldsProducers[i] != null) {
+        fieldsProducers[i] = fieldsProducers[i].getMergeInstance();
+      }
+      if (pointsReaders[i] != null) {
+        pointsReaders[i] = pointsReaders[i].getMergeInstance();
+      }
+      if (knnVectorsReaders[i] != null) {
+        knnVectorsReaders[i] = knnVectorsReaders[i].getMergeInstance();
+      }
+    }
+    return new MergeState(
+        docMaps,
+        segmentInfo,
+        mergeFieldInfos,
+        storedFieldsReaders,
+        termVectorsReaders,
+        normsProducers,
+        docValuesProducers,
+        fieldInfos,
+        liveDocs,
+        fieldsProducers,
+        pointsReaders,
+        knnVectorsReaders,
+        maxDocs,
+        infoStream,
+        intraMergeTaskExecutor,
+        needsIndexSort);
+  }
 }

lucene/core/src/java/org/apache/lucene/index/SegmentMerger.java

@@ -52,6 +52,7 @@ final class SegmentMerger {
 
   final MergeState mergeState;
   private final FieldInfos.Builder fieldInfosBuilder;
+  final Thread mergeStateCreationThread;
 
   // note, just like in codec apis Directory 'dir' is NOT the same as segmentInfo.dir!!
   SegmentMerger(
@@ -68,6 +69,7 @@ final class SegmentMerger {
           "IOContext.context should be MERGE; got: " + context.context());
     }
     mergeState = new MergeState(readers, segmentInfo, infoStream, intraMergeTaskExecutor);
+    mergeStateCreationThread = Thread.currentThread();
     directory = dir;
     this.codec = segmentInfo.getCodec();
     this.context = context;
@@ -99,6 +101,16 @@ final class SegmentMerger {
     return mergeState.segmentInfo.maxDoc() > 0;
   }
 
+  private MergeState mergeState() {
+    MergeState mergeState = this.mergeState;
+    if (Thread.currentThread() != mergeStateCreationThread) {
+      // Most merges, e.g. small merges, run in the same thread, so save the cost of pulling a clone
+      // in that case.
+      mergeState = mergeState.clone();
+    }
+    return mergeState;
+  }
+
   /**
    * Merges the readers into the directory passed to the constructor
    *
@@ -137,25 +149,15 @@ final class SegmentMerger {
 
     TaskExecutor taskExecutor = new TaskExecutor(mergeState.intraMergeTaskExecutor);
     List<Callable<Void>> mergingTasks = new ArrayList<>();
-    mergingTasks.add(
+    if (mergeState.mergeFieldInfos.hasNorms()) {
-        () -> {
+      mergeWithLogging(this::mergeNorms, segmentWriteState, segmentReadState, "norms", numMerged);
-          if (mergeState.mergeFieldInfos.hasNorms()) {
+    }
-            mergeWithLogging(
-                this::mergeNorms, segmentWriteState, segmentReadState, "norms", numMerged);
-          }
 
-          mergeWithLogging(
+    mergeWithLogging(this::mergeTerms, segmentWriteState, segmentReadState, "postings", numMerged);
-              this::mergeTerms, segmentWriteState, segmentReadState, "postings", numMerged);
-          return null;
-        });
 
     if (mergeState.mergeFieldInfos.hasDocValues()) {
-      mergingTasks.add(
+      mergeWithLogging(
-          () -> {
+          this::mergeDocValues, segmentWriteState, segmentReadState, "doc values", numMerged);
-            mergeWithLogging(
-                this::mergeDocValues, segmentWriteState, segmentReadState, "doc values", numMerged);
-            return null;
-          });
     }
 
     if (mergeState.mergeFieldInfos.hasPointValues()) {
@@ -201,6 +203,7 @@ final class SegmentMerger {
 
   private void mergeDocValues(
       SegmentWriteState segmentWriteState, SegmentReadState segmentReadState) throws IOException {
+    MergeState mergeState = mergeState();
     try (DocValuesConsumer consumer = codec.docValuesFormat().fieldsConsumer(segmentWriteState)) {
       consumer.merge(mergeState);
     }
@@ -208,6 +211,7 @@ final class SegmentMerger {
 
   private void mergePoints(SegmentWriteState segmentWriteState, SegmentReadState segmentReadState)
       throws IOException {
+    MergeState mergeState = mergeState();
     try (PointsWriter writer = codec.pointsFormat().fieldsWriter(segmentWriteState)) {
       writer.merge(mergeState);
     }
@@ -215,6 +219,7 @@ final class SegmentMerger {
 
   private void mergeNorms(SegmentWriteState segmentWriteState, SegmentReadState segmentReadState)
       throws IOException {
+    MergeState mergeState = mergeState();
     try (NormsConsumer consumer = codec.normsFormat().normsConsumer(segmentWriteState)) {
       consumer.merge(mergeState);
     }
@@ -222,6 +227,7 @@ final class SegmentMerger {
 
   private void mergeTerms(SegmentWriteState segmentWriteState, SegmentReadState segmentReadState)
       throws IOException {
+    MergeState mergeState = mergeState();
     try (NormsProducer norms =
         mergeState.mergeFieldInfos.hasNorms()
             ? codec.normsFormat().normsProducer(segmentReadState)
@@ -256,6 +262,7 @@ final class SegmentMerger {
    * @throws IOException if there is a low-level IO error
    */
   private int mergeFields() throws IOException {
+    MergeState mergeState = mergeState();
     try (StoredFieldsWriter fieldsWriter =
         codec.storedFieldsFormat().fieldsWriter(directory, mergeState.segmentInfo, context)) {
       return fieldsWriter.merge(mergeState);
@@ -268,6 +275,7 @@ final class SegmentMerger {
    * @throws IOException if there is a low-level IO error
    */
   private int mergeTermVectors() throws IOException {
+    MergeState mergeState = mergeState();
     try (TermVectorsWriter termVectorsWriter =
         codec.termVectorsFormat().vectorsWriter(directory, mergeState.segmentInfo, context)) {
       int numMerged = termVectorsWriter.merge(mergeState);
@@ -278,6 +286,7 @@ final class SegmentMerger {
 
   private void mergeVectorValues(
       SegmentWriteState segmentWriteState, SegmentReadState segmentReadState) throws IOException {
+    MergeState mergeState = mergeState();
     try (KnnVectorsWriter writer = codec.knnVectorsFormat().fieldsWriter(segmentWriteState)) {
       writer.merge(mergeState);
     }

lucene/core/src/test/org/apache/lucene/index/TestIndexWriterForceMerge.java

@@ -308,6 +308,7 @@ public class TestIndexWriterForceMerge extends LuceneTestCase {
     dir.close();
   }
 
+  @AwaitsFix(bugUrl = "https://github.com/apache/lucene/issues/13478")
   public void testMergePerField() throws IOException {
     IndexWriterConfig config = new IndexWriterConfig();
     ConcurrentMergeScheduler mergeScheduler =

lucene/test-framework/src/java/org/apache/lucene/tests/util/LuceneTestCase.java

@@ -94,6 +94,7 @@ import java.util.Random;
 import java.util.Set;
 import java.util.TimeZone;
 import java.util.TreeSet;
+import java.util.concurrent.Executor;
 import java.util.concurrent.ExecutorService;
 import java.util.concurrent.LinkedBlockingQueue;
 import java.util.concurrent.ThreadPoolExecutor;
@@ -943,10 +944,10 @@ public abstract class LuceneTestCase extends Assert {
     } else if (rarely(r)) {
       ConcurrentMergeScheduler cms;
       if (r.nextBoolean()) {
-        cms = new ConcurrentMergeScheduler();
+        cms = new TestConcurrentMergeScheduler();
       } else {
         cms =
-            new ConcurrentMergeScheduler() {
+            new TestConcurrentMergeScheduler() {
               @Override
               protected synchronized boolean maybeStall(MergeSource mergeSource) {
                 return true;
@@ -965,7 +966,8 @@ public abstract class LuceneTestCase extends Assert {
     } else {
       // Always use consistent settings, else CMS's dynamic (SSD or not)
       // defaults can change, hurting reproducibility:
-      ConcurrentMergeScheduler cms = new ConcurrentMergeScheduler();
+      ConcurrentMergeScheduler cms =
+          randomBoolean() ? new TestConcurrentMergeScheduler() : new ConcurrentMergeScheduler();
 
       // Only 1 thread can run at once (should maybe help reproducibility),
       // with up to 3 pending merges before segment-producing threads are
@@ -3292,4 +3294,17 @@ public abstract class LuceneTestCase extends Assert {
             .toList();
     return RandomPicks.randomFrom(random(), availableFormats);
   }
+
+  /**
+   * This is a test merge scheduler that will always use the intra merge executor to ensure we test
+   * it.
+   */
+  static class TestConcurrentMergeScheduler extends ConcurrentMergeScheduler {
+    @Override
+    public Executor getIntraMergeExecutor(MergePolicy.OneMerge merge) {
+      assert intraMergeExecutor != null : "scaledExecutor is not initialized";
+      // Always do the intra merge executor to ensure we test it
+      return intraMergeExecutor;
+    }
+  }
 }