TanYuxin-tyx commented on code in PR #22501: URL: https://github.com/apache/flink/pull/22501#discussion_r1189414413
########## flink-runtime/src/main/java/org/apache/flink/runtime/io/network/partition/hybrid/tiered/storage/SubpartitionCachedBuffer.java: ########## @@ -0,0 +1,193 @@ +/* + * 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.flink.runtime.io.network.partition.hybrid.tiered.storage; + +import org.apache.flink.core.memory.MemorySegment; +import org.apache.flink.core.memory.MemorySegmentFactory; +import org.apache.flink.runtime.io.network.buffer.Buffer; +import org.apache.flink.runtime.io.network.buffer.BufferBuilder; +import org.apache.flink.runtime.io.network.buffer.BufferConsumer; +import org.apache.flink.runtime.io.network.buffer.FreeingBufferRecycler; +import org.apache.flink.runtime.io.network.buffer.NetworkBuffer; +import org.apache.flink.runtime.io.network.partition.hybrid.tiered.common.TieredStorageSubpartitionId; + +import java.io.IOException; +import java.nio.ByteBuffer; +import java.util.Collections; +import java.util.LinkedList; +import java.util.List; +import java.util.Optional; +import java.util.Queue; +import java.util.function.BiConsumer; + +import static org.apache.flink.util.Preconditions.checkArgument; +import static org.apache.flink.util.Preconditions.checkNotNull; +import static org.apache.flink.util.Preconditions.checkState; + +/** + * {@link SubpartitionCachedBuffer} accumulates the records in a subpartition. + * + * <p>Note that {@link #setup} need an argument of buffer flush listener to accept the finished + * accumulated buffers. + */ +public class SubpartitionCachedBuffer { + + private final TieredStorageSubpartitionId subpartitionId; + + private final int bufferSize; + + private final TieredStorageMemoryManager storageMemoryManager; + + private final CacheFlushManager cacheFlushManager; + + private BiConsumer<TieredStorageSubpartitionId, List<Buffer>> bufferFlusher; + + // Not guarded by lock because it is expected only accessed from task's main thread. + private final Queue<BufferBuilder> unfinishedBuffers = new LinkedList<>(); + + public SubpartitionCachedBuffer( + TieredStorageSubpartitionId subpartitionId, + int bufferSize, + TieredStorageMemoryManager storageMemoryManager, + CacheFlushManager cacheFlushManager) { + this.subpartitionId = subpartitionId; + this.bufferSize = bufferSize; + this.storageMemoryManager = storageMemoryManager; + this.cacheFlushManager = cacheFlushManager; + } + + public void setup(BiConsumer<TieredStorageSubpartitionId, List<Buffer>> bufferFlusher) { + this.bufferFlusher = bufferFlusher; + } + + // ------------------------------------------------------------------------ + // Called by HashBasedCachedBuffer + // ------------------------------------------------------------------------ + + public void append(ByteBuffer record, Buffer.DataType dataType) + throws IOException, InterruptedException { + if (dataType.isEvent()) { + writeEvent(record, dataType); + } else { + writeRecord(record, dataType); + } + } + + public void close() { + checkState(unfinishedBuffers.isEmpty(), "There are unfinished buffers."); + } + + // ------------------------------------------------------------------------ + // Internal Methods + // ------------------------------------------------------------------------ + + private void writeEvent(ByteBuffer event, Buffer.DataType dataType) { + checkArgument(dataType.isEvent()); + + // each Event must take an exclusive buffer + finishCurrentWritingBufferIfNotEmpty(); + + // store Events in adhoc heap segments, for network memory efficiency + MemorySegment data = MemorySegmentFactory.wrap(event.array()); + addFinishedBuffer( + // Note that the buffer recycler should be replaced before writing to the tiers. + new NetworkBuffer(data, FreeingBufferRecycler.INSTANCE, dataType, data.size())); + } + + private void writeRecord(ByteBuffer record, Buffer.DataType dataType) { + checkArgument(!dataType.isEvent()); + + ensureCapacityForRecord(record); + + writeRecord(record); + } + + private void ensureCapacityForRecord(ByteBuffer record) { + final int numRecordBytes = record.remaining(); + int availableBytes = + Optional.ofNullable(unfinishedBuffers.peek()) + .map( + currentWritingBuffer -> + currentWritingBuffer.getWritableBytes() + + bufferSize * (unfinishedBuffers.size() - 1)) + .orElse(0); + + while (availableBytes < numRecordBytes) { + // request unfinished buffer. + BufferBuilder bufferBuilder = requestBufferFromPool(); + unfinishedBuffers.add(bufferBuilder); + availableBytes += bufferSize; + } + } + + private void writeRecord(ByteBuffer record) { + while (record.hasRemaining()) { + BufferBuilder currentWritingBuffer = + checkNotNull( + unfinishedBuffers.peek(), "Expect enough capacity for the record."); + currentWritingBuffer.append(record); + if (currentWritingBuffer.isFull()) { + finishCurrentWritingBuffer(); + } + } + } + + private void finishCurrentWritingBufferIfNotEmpty() { + BufferBuilder currentWritingBuffer = unfinishedBuffers.peek(); + if (currentWritingBuffer == null || currentWritingBuffer.getWritableBytes() == bufferSize) { + return; + } + + finishCurrentWritingBuffer(); + } + + private void finishCurrentWritingBuffer() { + BufferBuilder currentWritingBuffer = unfinishedBuffers.poll(); + if (currentWritingBuffer == null) { + return; + } + currentWritingBuffer.finish(); + BufferConsumer bufferConsumer = currentWritingBuffer.createBufferConsumerFromBeginning(); + Buffer buffer = bufferConsumer.build(); + currentWritingBuffer.close(); + bufferConsumer.close(); + addFinishedBuffer( + new NetworkBuffer( + buffer.getMemorySegment(), + // Note that the buffer recycler should be replaced before writing to the + // tiers. + FreeingBufferRecycler.INSTANCE, + buffer.getDataType(), + buffer.getSize())); + } + + private void addFinishedBuffer(Buffer finishedBuffer) { + bufferFlusher.accept(subpartitionId, Collections.singletonList(finishedBuffer)); + } + + public BufferBuilder requestBufferFromPool() { + MemorySegment segment = storageMemoryManager.requestBufferInAccumulator(); + cacheFlushManager.checkNeedTriggerFlushCachedBuffers(); Review Comment: After requesting a new buffer from the memory manager, we should check whether it is needed to flush the cached buffers to release buffers. This may include flushing the cached buffers to disk or remote storage. -- 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. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
