yupeng9 commented on a change in pull request #6113:
URL: https://github.com/apache/incubator-pinot/pull/6113#discussion_r501128615
##########
File path:
pinot-core/src/main/java/org/apache/pinot/core/data/manager/realtime/RealtimeTableDataManager.java
##########
@@ -266,14 +291,59 @@ public void addSegment(String segmentName, TableConfig
tableConfig, IndexLoading
manager =
new LLRealtimeSegmentDataManager(realtimeSegmentZKMetadata,
tableConfig, this, _indexDir.getAbsolutePath(),
indexLoadingConfig, schema, llcSegmentName,
_partitionIdToSemaphoreMap.get(streamPartitionId),
- _serverMetrics);
+ _serverMetrics, _upsertMetadataTableManager);
}
_logger.info("Initialize RealtimeSegmentDataManager - " + segmentName);
_segmentDataManagerMap.put(segmentName, manager);
_serverMetrics.addValueToTableGauge(_tableNameWithType,
ServerGauge.SEGMENT_COUNT, 1L);
}
}
+ private boolean isUpsertEnabled() {
+ return _upsertMode != null && _upsertMode == UpsertConfig.Mode.FULL ||
_upsertMode == UpsertConfig.Mode.PARTIAL;
+ }
+
+ @Override
+ public void addSegment(ImmutableSegment immutableSegment) {
+ if(isUpsertEnabled()) {
+ handleUpsert(immutableSegment);
+ }
+ super.addSegment(immutableSegment);
+ }
+
+ private void handleUpsert(ImmutableSegment immutableSegment) {
+ Preconditions.checkArgument(!_primaryKeyColumns.isEmpty(), "the primary
key columns cannot be empty");
+ Map<String, PinotSegmentColumnReader> columnToReaderMap = new HashMap<>();
+ for (String primaryKeyColumn : _primaryKeyColumns) {
+ columnToReaderMap.put(primaryKeyColumn, new
PinotSegmentColumnReader(immutableSegment, primaryKeyColumn));
+ }
+ columnToReaderMap.put(_timeColumnName, new
PinotSegmentColumnReader(immutableSegment, _timeColumnName));
+ int numTotalDocs = immutableSegment.getSegmentMetadata().getTotalDocs();
+
+ // upsert metadata of the current segment
+ Map<PrimaryKey, RecordLocation> primaryKeyIndex = new HashMap<>();
Review comment:
Agreed on being careful about memory usage. I discussed this in the
design doc, and my estimate of the memory usage shall be < 2GB.
I do plan a followup PR of adding metrics of the memory usage.
##########
File path:
pinot-core/src/main/java/org/apache/pinot/core/upsert/UpsertProcessorUtil.java
##########
@@ -0,0 +1,69 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+package org.apache.pinot.core.upsert;
+
+import java.util.Map;
+import org.apache.pinot.core.realtime.impl.ThreadSafeMutableRoaringBitmap;
+import org.apache.pinot.spi.data.readers.PrimaryKey;
+import org.slf4j.Logger;
+import org.slf4j.LoggerFactory;
+
+
+public final class UpsertProcessorUtil {
+ private static final Logger LOGGER =
LoggerFactory.getLogger(UpsertProcessorUtil.class);
+
+ private UpsertProcessorUtil() {
+ }
+
+ public static void handleUpsert(PrimaryKey primaryKey, long timestamp,
String segmentName, int docId, int partitionId,
+ Map<PrimaryKey, RecordLocation> primaryKeyIndex,
ThreadSafeMutableRoaringBitmap validDocIndex,
+ TableUpsertMetadataManager upsertMetadataTableManager) {
+ RecordLocation location = new RecordLocation(segmentName, docId,
timestamp);
+ // check local primary key index first
Review comment:
I think all segments are in-memory? Changed it to for mutable segment.
##########
File path:
pinot-core/src/main/java/org/apache/pinot/core/upsert/TableUpsertMetadataManager.java
##########
@@ -0,0 +1,73 @@
+/**
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements. See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership. The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License. You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing,
+ * software distributed under the License is distributed on an
+ * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+ * KIND, either express or implied. See the License for the
+ * specific language governing permissions and limitations
+ * under the License.
+ */
+package org.apache.pinot.core.upsert;
+
+import java.util.Map;
+import java.util.concurrent.ConcurrentHashMap;
+import javax.annotation.concurrent.ThreadSafe;
+import org.apache.pinot.core.realtime.impl.ThreadSafeMutableRoaringBitmap;
+import org.apache.pinot.spi.data.readers.PrimaryKey;
+
+/**
+ * The manager of the upsert metadata of a table.
+ */
+@ThreadSafe
+public class TableUpsertMetadataManager {
+ private final Map<Integer, PartitionUpsertMetadataManager>
_partitionMetadataManagerMap = new ConcurrentHashMap<>();
Review comment:
No, this is per server, so it tracks local segment only. And there is at
most one replica per server
##########
File path:
pinot-core/src/main/java/org/apache/pinot/core/data/manager/realtime/RealtimeTableDataManager.java
##########
@@ -266,14 +291,59 @@ public void addSegment(String segmentName, TableConfig
tableConfig, IndexLoading
manager =
new LLRealtimeSegmentDataManager(realtimeSegmentZKMetadata,
tableConfig, this, _indexDir.getAbsolutePath(),
indexLoadingConfig, schema, llcSegmentName,
_partitionIdToSemaphoreMap.get(streamPartitionId),
- _serverMetrics);
+ _serverMetrics, _upsertMetadataTableManager);
}
_logger.info("Initialize RealtimeSegmentDataManager - " + segmentName);
_segmentDataManagerMap.put(segmentName, manager);
_serverMetrics.addValueToTableGauge(_tableNameWithType,
ServerGauge.SEGMENT_COUNT, 1L);
}
}
+ private boolean isUpsertEnabled() {
+ return _upsertMode != null && _upsertMode == UpsertConfig.Mode.FULL ||
_upsertMode == UpsertConfig.Mode.PARTIAL;
+ }
+
+ @Override
+ public void addSegment(ImmutableSegment immutableSegment) {
+ if(isUpsertEnabled()) {
+ handleUpsert(immutableSegment);
+ }
+ super.addSegment(immutableSegment);
+ }
+
+ private void handleUpsert(ImmutableSegment immutableSegment) {
+ Preconditions.checkArgument(!_primaryKeyColumns.isEmpty(), "the primary
key columns cannot be empty");
+ Map<String, PinotSegmentColumnReader> columnToReaderMap = new HashMap<>();
+ for (String primaryKeyColumn : _primaryKeyColumns) {
+ columnToReaderMap.put(primaryKeyColumn, new
PinotSegmentColumnReader(immutableSegment, primaryKeyColumn));
+ }
+ columnToReaderMap.put(_timeColumnName, new
PinotSegmentColumnReader(immutableSegment, _timeColumnName));
+ int numTotalDocs = immutableSegment.getSegmentMetadata().getTotalDocs();
+
+ // upsert metadata of the current segment
+ Map<PrimaryKey, RecordLocation> primaryKeyIndex = new HashMap<>();
+ ThreadSafeMutableRoaringBitmap validDocIndex = new
ThreadSafeMutableRoaringBitmap();
+
+ String segmentName = immutableSegment.getSegmentName();
+ int partitionId = new
LLCSegmentName(immutableSegment.getSegmentName()).getPartitionId();
+ for (int docId = 0; docId < numTotalDocs; docId++) {
Review comment:
That's a good point. I think some optimization on batch loading the
columns can be considered in future PRs. In fact, this may improve other index
loading code path such as the null vector loading etc.
In the initial version, we can build the foundation for the functionality
first, and later via profiling we will have better understanding of the
performance bottlenecks, and we can improve accordingly.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: commits-unsubscr...@pinot.apache.org
For additional commands, e-mail: commits-h...@pinot.apache.org
