jamesyfshao commented on a change in pull request #5394:
URL: https://github.com/apache/incubator-pinot/pull/5394#discussion_r446715531



##########
File path: 
pinot-core/src/main/java/org/apache/pinot/core/data/manager/callback/DataManagerCallback.java
##########
@@ -0,0 +1,106 @@
+/**
+ * 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.data.manager.callback;
+
+import org.apache.pinot.core.data.manager.SegmentDataManager;
+import org.apache.pinot.spi.annotations.InterfaceStability;
+import org.apache.pinot.spi.data.readers.GenericRow;
+import org.apache.pinot.spi.stream.MessageBatch;
+import org.apache.pinot.spi.stream.StreamPartitionMsgOffset;
+
+import java.io.IOException;
+
+/**
+ * Component inject to {@link 
org.apache.pinot.core.data.manager.SegmentDataManager} for handling extra logic 
for
+ * upsert-enabled pinot ingestion mode.
+ */
+@InterfaceStability.Evolving
+public interface DataManagerCallback {
+
+  void init() throws IOException;
+
+  /**
+   * create a {@link IndexSegmentCallback} object to allow SegmentDataManager 
to create proper IndexSegment that supports
+   * either append/upsert mode
+   *
+   * In append-tables callback, this method will create a 
DefaultIndexSegmentCallback
+   * In upsert-tables callback, this method will create a 
UpsertDataManagerCallbackImpl
+   *
+   * The {@link IndexSegmentCallback} will be used in the constructor of 
IndexSegment
+   */
+  IndexSegmentCallback getIndexSegmentCallback();
+
+  /**
+   * process the row after transformation in 
LLRealtimeSegmentDataManager.processStreamEvents(...) method
+   * it happens after the GenericRow has been transformed by RecordTransformer 
and before it is indexed by
+   * IndexSegmentImpl, to ensure we can provide other necessary data to the 
segment metadata.
+   *
+   * In append-tables callback, this method will do nothing
+   * In upsert-tables callback, this method will attach the offset object into 
the GenericRow object.
+   *
+   * The reason we need this particular logic is that in upsert table, we need 
to add offset data to the physical data
+   * this will help us to apply the update events from key coordinator to 
upsert table correctly as the offset
+   * is used as the index to identify which record's virtual column we want to 
update
+   *
+   * @param row the row of newly ingested and transformed data from upstream
+   * @param offset the offset of this particular pinot record
+   */
+  void processTransformedRow(GenericRow row, StreamPartitionMsgOffset offset);
+
+  /**
+   * process the row after indexing in 
LLRealtimeSegmentDataManager.processStreamEvents(...) method
+   * it happens after the MutableSegmentImpl has done the indexing of the 
current row in its physical storage
+   *
+   * In append-tables callback, this method will do nothing
+   * In upsert-tables callback, this method will emit an event to the message 
queue that will deliver the event to
+   * key coordinator.
+   *
+   * This method ensures that we can emit the metadata for an new entry that 
pinot just indexed to its internal storage
+   * and let key coordinator to be able to consume those events to process the 
updates correctly
+   *
+   * @param row the row we just index in the current segment
+   * @param offset the offset associated with the current row
+   */
+  void postIndexProcessing(GenericRow row, StreamPartitionMsgOffset offset);
+
+  /**
+   * perform any necessary finalize operation after the consumer loop finished 
in LLRealtimeSegmentDataManager.consumeLoop(...)
+   * method. It happens after the consumerloop exits the loop and reached the 
end criteria.
+   *
+   * In append-tables callback, this method will do nothing
+   * In upsert-tables callback, this method will flush the queue producer to 
ensure all pending messages are deliverd
+   * to the queue between pinot server and pinot key-coordinator
+   *
+   * this method will ensure that pinot server can send all events to key 
coordinator eventually before a segment
+   * is committed. If this does not happen we might lose data in case of 
machine failure.
+   */
+  void postConsumeLoop();

Review comment:
       Update the method name (onConsumptionStoppedOrEndReached) to indicate 
that this method is supposed to happen during the consumption 
stop/end-condition reached. Right now it is fine to call this multiple times if 
we got catch up call. The purpose of the method is indicated at the method doc: 
(ensure that pinot server can send all events to key coordinator eventually 
before a segment is committed). Do you think the new naming would be clear 
enough to a reader or you think we should add more to the method documentation?




----------------------------------------------------------------
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

Reply via email to