mon: add diagnose information for MemoryNotEnoughException (#17578)

Author: shizy818

iotdb-core/datanode/src/main/java/org/apache/iotdb/db/queryengine/execution/fragment/FragmentInstanceContext.java

@@ -934,7 +934,7 @@ public void releaseResourceWhenAllDriversAreClosed() {
    */
   private void releaseTVListOwnedByQuery() {
     for (TVList tvList : tvListSet) {
-      long tvListRamSize = tvList.calculateRamSize();
+      long tvListRamSize = tvList.calculateRamSize().getRamSize();
       tvList.lockQueryList();
       Set<QueryContext> queryContextSet = tvList.getQueryContextSet();
       try {

iotdb-core/datanode/src/main/java/org/apache/iotdb/db/schemaengine/schemaregion/utils/ResourceByPathUtils.java

@@ -19,6 +19,7 @@
 
 package org.apache.iotdb.db.schemaengine.schemaregion.utils;
 
+import org.apache.iotdb.calc.exception.MemoryNotEnoughException;
 import org.apache.iotdb.calc.exception.QueryProcessException;
 import org.apache.iotdb.calc.plan.planner.memory.MemoryReservationManager;
 import org.apache.iotdb.commons.path.AlignedFullPath;
@@ -155,7 +156,7 @@ protected Map<TVList, Integer> prepareTvListMapForQuery(
     // mutable tvlist
     TVList list = memChunk.getWorkingTVList();
     TVList cloneList = null;
-    long tvListRamSize = list.calculateRamSize();
+    TVList.RamInfo listRamInfo = list.calculateRamSize();
     list.lockQueryList();
     try {
       if (copyTimeFilter != null
@@ -196,8 +197,8 @@ protected Map<TVList, Integer> prepareTvListMapForQuery(
           if (firstQuery instanceof FragmentInstanceContext) {
             MemoryReservationManager memoryReservationManager =
                 ((FragmentInstanceContext) firstQuery).getMemoryReservationContext();
-            memoryReservationManager.reserveMemoryCumulatively(tvListRamSize);
-            list.setReservedMemoryBytes(tvListRamSize);
+            memoryReservationManager.reserveMemoryCumulatively(listRamInfo.getRamSize());
+            list.setReservedMemoryBytes(listRamInfo.getRamSize());
           }
           list.setOwnerQuery(firstQuery);
 
@@ -207,6 +208,14 @@ protected Map<TVList, Integer> prepareTvListMapForQuery(
           tvListQueryMap.put(cloneList, cloneList.rowCount());
         }
       }
+    } catch (MemoryNotEnoughException ex) {
+      LOGGER.warn(
+          "Failed to reserve memory for TVList: ramSize {}, timestampsSize {}, arrayMemCost {}, rowCount {}, dataTypes {}",
+          listRamInfo.getRamSize(),
+          listRamInfo.getTimestampsSize(),
+          listRamInfo.getArrayMemCost(),
+          listRamInfo.getRowCount(),
+          listRamInfo.getDataTypes());
     } finally {
       list.unlockQueryList();
     }

iotdb-core/datanode/src/main/java/org/apache/iotdb/db/storageengine/dataregion/memtable/AbstractWritableMemChunk.java

@@ -103,7 +103,7 @@ protected void maybeReleaseTvList(TVList tvList) {
   }
 
   private void tryReleaseTvList(TVList tvList) {
-    long tvListRamSize = tvList.calculateRamSize();
+    long tvListRamSize = tvList.calculateRamSize().getRamSize();
     tvList.lockQueryList();
     try {
       if (tvList.getQueryContextSet().isEmpty()) {

iotdb-core/datanode/src/main/java/org/apache/iotdb/db/storageengine/dataregion/memtable/AlignedReadOnlyMemChunk.java

@@ -127,7 +127,7 @@ public void sortTvLists() {
         // We must update queryRowCount here, otherwise, it may be used later to build
         // BitMaps, causing bitmap array size mismatch and possible out of bound.
         entry.setValue(alignedTvList.sort());
-        long alignedTvListRamSize = alignedTvList.calculateRamSize();
+        long alignedTvListRamSize = alignedTvList.calculateRamSize().getRamSize();
         alignedTvList.lockQueryList();
         try {
           FragmentInstanceContext ownerQuery =
@@ -384,7 +384,7 @@ public IPointReader getPointReader() {
       int queryLength = entry.getValue();
       if (!alignedTvList.isSorted() && queryLength > alignedTvList.seqRowCount()) {
         entry.setValue(alignedTvList.sort());
-        long alignedTvListRamSize = alignedTvList.calculateRamSize();
+        long alignedTvListRamSize = alignedTvList.calculateRamSize().getRamSize();
         alignedTvList.lockQueryList();
         try {
           FragmentInstanceContext ownerQuery =

iotdb-core/datanode/src/main/java/org/apache/iotdb/db/storageengine/dataregion/memtable/ReadOnlyMemChunk.java

@@ -137,7 +137,7 @@ public void sortTvLists() {
       int queryRowCount = entry.getValue();
       if (!tvList.isSorted() && queryRowCount > tvList.seqRowCount()) {
         entry.setValue(tvList.sort());
-        long tvListRamSize = tvList.calculateRamSize();
+        long tvListRamSize = tvList.calculateRamSize().getRamSize();
         tvList.lockQueryList();
         try {
           FragmentInstanceContext ownerQuery = (FragmentInstanceContext) tvList.getOwnerQuery();
@@ -294,7 +294,7 @@ public IPointReader getPointReader() {
       int queryLength = entry.getValue();
       if (!tvList.isSorted() && queryLength > tvList.seqRowCount()) {
         entry.setValue(tvList.sort());
-        long tvListRamSize = tvList.calculateRamSize();
+        long tvListRamSize = tvList.calculateRamSize().getRamSize();
         tvList.lockQueryList();
         try {
           FragmentInstanceContext ownerQuery = (FragmentInstanceContext) tvList.getOwnerQuery();

iotdb-core/datanode/src/main/java/org/apache/iotdb/db/utils/datastructure/AlignedTVList.java

@@ -1064,8 +1064,9 @@ public TSDataType getDataType() {
   }
 
   @Override
-  public synchronized long calculateRamSize() {
-    return timestamps.size() * alignedTvListArrayMemCost();
+  public synchronized RamInfo calculateRamSize() {
+    return new RamInfo(
+        timestamps.size(), alignedTvListArrayMemCost(), rowCount, new ArrayList<>(dataTypes));
   }
 
   /**

iotdb-core/datanode/src/main/java/org/apache/iotdb/db/utils/datastructure/TVList.java

@@ -28,6 +28,7 @@
 import org.apache.iotdb.db.storageengine.rescon.memory.PrimitiveArrayManager;
 import org.apache.iotdb.db.utils.MathUtils;
 
+import com.google.common.collect.ImmutableList;
 import org.apache.tsfile.enums.TSDataType;
 import org.apache.tsfile.file.metadata.enums.TSEncoding;
 import org.apache.tsfile.read.TimeValuePair;
@@ -61,6 +62,42 @@
 
 public abstract class TVList implements WALEntryValue {
   protected static final String ERR_DATATYPE_NOT_CONSISTENT = "DataType not consistent";
+
+  public static class RamInfo {
+    private final int timestampsSize;
+    private final long arrayMemCost;
+    private final int rowCount;
+    private final List<TSDataType> dataTypes;
+
+    public RamInfo(
+        int timestampCount, long arrayMemCost, int rowCount, List<TSDataType> dataTypes) {
+      this.timestampsSize = timestampCount;
+      this.rowCount = rowCount;
+      this.arrayMemCost = arrayMemCost;
+      this.dataTypes = dataTypes;
+    }
+
+    public long getRamSize() {
+      return timestampsSize * arrayMemCost;
+    }
+
+    public int getTimestampsSize() {
+      return timestampsSize;
+    }
+
+    public int getRowCount() {
+      return rowCount;
+    }
+
+    public long getArrayMemCost() {
+      return arrayMemCost;
+    }
+
+    public List<TSDataType> getDataTypes() {
+      return dataTypes;
+    }
+  }
+
   // list of timestamp array, add 1 when expanded -> data point timestamp array
   // index relation: arrayIndex -> elementIndex
   protected List<long[]> timestamps;
@@ -165,8 +202,9 @@ public static long tvListArrayMemCost(TSDataType type) {
     return size;
   }
 
-  public synchronized long calculateRamSize() {
-    return timestamps.size() * tvListArrayMemCost();
+  public synchronized RamInfo calculateRamSize() {
+    return new RamInfo(
+        timestamps.size(), tvListArrayMemCost(), rowCount, ImmutableList.of(getDataType()));
   }
 
   public synchronized boolean isSorted() {

iotdb-core/datanode/src/test/java/org/apache/iotdb/db/queryengine/execution/fragment/FragmentInstanceExecutionTest.java

@@ -146,7 +146,7 @@ public void testTVListOwnerTransfer() throws InterruptedException {
         // mock flush's behavior
         fragmentInstanceContext1
             .getMemoryReservationContext()
-            .reserveMemoryCumulatively(tvList.calculateRamSize());
+            .reserveMemoryCumulatively(tvList.calculateRamSize().getRamSize());
         tvList.setOwnerQuery(fragmentInstanceContext1);
 
         fragmentInstanceContext1.decrementNumOfUnClosedDriver();
@@ -226,7 +226,7 @@ public void testTVListCloneForQuery() {
         pointReader.nextTimeValuePair();
       }
       assertTrue(tvList.isSorted());
-      assertEquals(tvList.calculateRamSize(), tvList.getReservedMemoryBytes());
+      assertEquals(tvList.calculateRamSize().getRamSize(), tvList.getReservedMemoryBytes());
     } catch (QueryProcessException
         | IOException
         | MetadataException

iotdb-core/datanode/src/test/java/org/apache/iotdb/db/storageengine/dataregion/memtable/PrimitiveMemTableTest.java

@@ -748,7 +748,7 @@ public void testReleaseWithNotEnoughMemory() throws CpuNotEnoughException {
         Mockito.mock(MemoryReservationManager.class);
     Mockito.doThrow(new MemoryNotEnoughException(""))
         .when(memoryReservationManager)
-        .reserveMemoryCumulatively(list.calculateRamSize());
+        .reserveMemoryCumulatively(list.calculateRamSize().getRamSize());
 
     // create FragmentInstanceId
     QueryId queryId = new QueryId("stub_query");