umamaheswararao commented on code in PR #3504: URL: https://github.com/apache/ozone/pull/3504#discussion_r898492701
########## hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/ec/reconstruction/ECReconstructionCoordinator.java: ########## @@ -0,0 +1,417 @@ +/* + * 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 + * <p> + * http://www.apache.org/licenses/LICENSE-2.0 + * <p> + * 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.hadoop.ozone.container.ec.reconstruction; + +import com.google.common.base.Preconditions; +import com.google.common.collect.ImmutableList; +import com.google.common.util.concurrent.ThreadFactoryBuilder; +import org.apache.hadoop.hdds.client.BlockID; +import org.apache.hadoop.hdds.client.ECReplicationConfig; +import org.apache.hadoop.hdds.conf.ConfigurationSource; +import org.apache.hadoop.hdds.protocol.DatanodeDetails; +import org.apache.hadoop.hdds.protocol.datanode.proto.ContainerProtos; +import org.apache.hadoop.hdds.scm.ByteStringConversion; +import org.apache.hadoop.hdds.scm.OzoneClientConfig; +import org.apache.hadoop.hdds.scm.pipeline.Pipeline; +import org.apache.hadoop.hdds.scm.pipeline.PipelineID; +import org.apache.hadoop.hdds.scm.storage.BlockLocationInfo; +import org.apache.hadoop.hdds.scm.storage.BufferPool; +import org.apache.hadoop.hdds.scm.storage.ECBlockOutputStream; +import org.apache.hadoop.hdds.security.x509.certificate.client.CertificateClient; +import org.apache.hadoop.io.ByteBufferPool; +import org.apache.hadoop.io.ElasticByteBufferPool; +import org.apache.hadoop.ozone.OzoneConsts; +import org.apache.hadoop.ozone.client.io.BlockInputStreamFactory; +import org.apache.hadoop.ozone.client.io.BlockInputStreamFactoryImpl; +import org.apache.hadoop.ozone.client.io.ECBlockInputStreamProxy; +import org.apache.hadoop.ozone.client.io.ECBlockReconstructedStripeInputStream; +import org.apache.hadoop.ozone.container.common.helpers.BlockData; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.io.Closeable; +import java.io.IOException; +import java.nio.ByteBuffer; +import java.util.ArrayList; +import java.util.HashMap; +import java.util.Iterator; +import java.util.List; +import java.util.Map; +import java.util.Set; +import java.util.SortedMap; +import java.util.TreeMap; +import java.util.concurrent.CompletableFuture; +import java.util.concurrent.ExecutionException; +import java.util.concurrent.ExecutorService; +import java.util.concurrent.SynchronousQueue; +import java.util.concurrent.ThreadPoolExecutor; +import java.util.concurrent.TimeUnit; +import java.util.stream.Collectors; + +/** + * The Coordinator implements the main flow of reconstructing + * missing container replicas. + * <p> + * For a container reconstruction task, the main flow is: + * - ListBlock from all healthy replicas + * - calculate effective block group len for all blocks + * - for each block + * - build a ReconstructInputStream to read healthy chunks + * - build a ECBlockOutputStream to write out decoded chunks + * - for each stripe + * - use ReconstructInputStream.readStripe to decode missing chunks + * - use ECBlockOutputStream.write to write decoded chunks to TargetDNs + * - PutBlock + * - CloseContainer + */ +public class ECReconstructionCoordinator implements Closeable { + + static final Logger LOG = + LoggerFactory.getLogger(ECReconstructionCoordinator.class); + + private static final int EC_RECONSTRUCT_STRIPE_READ_POOL_MIN_SIZE = 3; + + private final ECContainerOperationClient containerOperationClient; + + private final ConfigurationSource config; + + private final ByteBufferPool byteBufferPool; + + private ExecutorService ecReconstructExecutor; + + private BlockInputStreamFactory blockInputStreamFactory; + + public ECReconstructionCoordinator(ECContainerOperationClient containerClient, + ConfigurationSource conf, ByteBufferPool byteBufferPool, + ExecutorService reconstructExecutor, + BlockInputStreamFactory streamFactory) { + this.containerOperationClient = containerClient; + this.config = conf; + this.byteBufferPool = byteBufferPool; + this.blockInputStreamFactory = streamFactory; + this.ecReconstructExecutor = reconstructExecutor; + } + + public ECReconstructionCoordinator(ConfigurationSource conf, + CertificateClient certificateClient) throws IOException { + this(new ECContainerOperationClient(conf, certificateClient), conf, + new ElasticByteBufferPool(), null, null); + this.ecReconstructExecutor = + new ThreadPoolExecutor(EC_RECONSTRUCT_STRIPE_READ_POOL_MIN_SIZE, + config.getObject(OzoneClientConfig.class) + .getEcReconstructStripeReadPoolLimit(), 60, TimeUnit.SECONDS, + new SynchronousQueue<>(), new ThreadFactoryBuilder() + .setNameFormat("ec-reconstruct-reader-TID-%d").build(), + new ThreadPoolExecutor.CallerRunsPolicy()); + this.blockInputStreamFactory = BlockInputStreamFactoryImpl + .getInstance(byteBufferPool, () -> ecReconstructExecutor); + } + + public void reconstructECContainerGroup(long containerID, + ECReplicationConfig repConfig, + SortedMap<Integer, DatanodeDetails> sourceNodeMap, + SortedMap<Integer, DatanodeDetails> targetNodeMap) throws IOException { + + Pipeline pipeline = rebuildInputPipeline(repConfig, sourceNodeMap); + + SortedMap<Long, BlockData[]> blockDataMap = + getBlockDataMap(containerID, repConfig, sourceNodeMap); + + SortedMap<Long, BlockLocationInfo> blockLocationInfoMap = + calcBlockLocationInfoMap(containerID, blockDataMap, pipeline); + + // 1. create target recovering containers. + Set<Map.Entry<Integer, DatanodeDetails>> targetIndexDns = + targetNodeMap.entrySet(); + Iterator<Map.Entry<Integer, DatanodeDetails>> iterator = + targetIndexDns.iterator(); + while (iterator.hasNext()) { + Map.Entry<Integer, DatanodeDetails> indexDnPair = iterator.next(); + this.containerOperationClient + .createRecoveringContainer(containerID, indexDnPair.getValue(), + repConfig, null, indexDnPair.getKey()); + } + + // 2. Reconstruct and transfer to targets + for (BlockLocationInfo blockLocationInfo : blockLocationInfoMap.values()) { + reconstructECBlockGroup(blockLocationInfo, repConfig, targetNodeMap); + } + + // 3. Close containers + iterator = targetIndexDns.iterator(); + while (iterator.hasNext()) { + DatanodeDetails dn = iterator.next().getValue(); + this.containerOperationClient + .closeContainer(containerID, dn, repConfig, null); + } + + } + + void reconstructECBlockGroup(BlockLocationInfo blockLocationInfo, + ECReplicationConfig repConfig, + SortedMap<Integer, DatanodeDetails> targetMap) + throws IOException { + long safeBlockGroupLength = blockLocationInfo.getLength(); + List<Integer> missingContainerIndexes = + targetMap.keySet().stream().collect(Collectors.toList()); + + // calculate the real missing block indexes + int dataLocs = ECBlockInputStreamProxy + .expectedDataLocations(repConfig, safeBlockGroupLength); + List<Integer> toReconstructIndexes = new ArrayList<>(); + for (int i = 0; i < missingContainerIndexes.size(); i++) { + Integer index = missingContainerIndexes.get(i); + if (index <= dataLocs || index > repConfig.getData()) { + toReconstructIndexes.add(index); + } + // else padded indexes. + } + + // Looks like we don't need to reconstruct any missing blocks in this block + // group. The reason for this should be block group had only padding blocks + // in the missing locations. + if (toReconstructIndexes.size() == 0) { + if (LOG.isDebugEnabled()) { + LOG.debug("Skipping the reconstruction for the block: " + + blockLocationInfo.getBlockID() + ". In the missing locations: " + + missingContainerIndexes + + ", this block group has only padded blocks."); + } + return; + } + + try (ECBlockReconstructedStripeInputStream sis + = new ECBlockReconstructedStripeInputStream( + repConfig, blockLocationInfo, true, + this.containerOperationClient.getXceiverClientManager(), null, + this.blockInputStreamFactory, byteBufferPool, + this.ecReconstructExecutor)) { + + ECBlockOutputStream[] targetBlockStreams = + new ECBlockOutputStream[toReconstructIndexes.size()]; + ByteBuffer[] bufs = new ByteBuffer[toReconstructIndexes.size()]; + for (int i = 0; i < toReconstructIndexes.size(); i++) { + OzoneClientConfig configuration = new OzoneClientConfig(); + // TODO: Let's avoid unnecessary bufferPool creation for + BufferPool bufferPool = + new BufferPool(configuration.getStreamBufferSize(), + (int) (configuration.getStreamBufferMaxSize() / configuration + .getStreamBufferSize()), + ByteStringConversion.createByteBufferConversion(false)); Review Comment: Sorry, it was partially written todo, I updated it. ########## hadoop-hdds/container-service/src/main/java/org/apache/hadoop/ozone/container/ec/reconstruction/ECReconstructionCoordinator.java: ########## @@ -0,0 +1,417 @@ +/* + * 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 + * <p> + * http://www.apache.org/licenses/LICENSE-2.0 + * <p> + * 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.hadoop.ozone.container.ec.reconstruction; + +import com.google.common.base.Preconditions; +import com.google.common.collect.ImmutableList; +import com.google.common.util.concurrent.ThreadFactoryBuilder; +import org.apache.hadoop.hdds.client.BlockID; +import org.apache.hadoop.hdds.client.ECReplicationConfig; +import org.apache.hadoop.hdds.conf.ConfigurationSource; +import org.apache.hadoop.hdds.protocol.DatanodeDetails; +import org.apache.hadoop.hdds.protocol.datanode.proto.ContainerProtos; +import org.apache.hadoop.hdds.scm.ByteStringConversion; +import org.apache.hadoop.hdds.scm.OzoneClientConfig; +import org.apache.hadoop.hdds.scm.pipeline.Pipeline; +import org.apache.hadoop.hdds.scm.pipeline.PipelineID; +import org.apache.hadoop.hdds.scm.storage.BlockLocationInfo; +import org.apache.hadoop.hdds.scm.storage.BufferPool; +import org.apache.hadoop.hdds.scm.storage.ECBlockOutputStream; +import org.apache.hadoop.hdds.security.x509.certificate.client.CertificateClient; +import org.apache.hadoop.io.ByteBufferPool; +import org.apache.hadoop.io.ElasticByteBufferPool; +import org.apache.hadoop.ozone.OzoneConsts; +import org.apache.hadoop.ozone.client.io.BlockInputStreamFactory; +import org.apache.hadoop.ozone.client.io.BlockInputStreamFactoryImpl; +import org.apache.hadoop.ozone.client.io.ECBlockInputStreamProxy; +import org.apache.hadoop.ozone.client.io.ECBlockReconstructedStripeInputStream; +import org.apache.hadoop.ozone.container.common.helpers.BlockData; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +import java.io.Closeable; +import java.io.IOException; +import java.nio.ByteBuffer; +import java.util.ArrayList; +import java.util.HashMap; +import java.util.Iterator; +import java.util.List; +import java.util.Map; +import java.util.Set; +import java.util.SortedMap; +import java.util.TreeMap; +import java.util.concurrent.CompletableFuture; +import java.util.concurrent.ExecutionException; +import java.util.concurrent.ExecutorService; +import java.util.concurrent.SynchronousQueue; +import java.util.concurrent.ThreadPoolExecutor; +import java.util.concurrent.TimeUnit; +import java.util.stream.Collectors; + +/** + * The Coordinator implements the main flow of reconstructing + * missing container replicas. + * <p> + * For a container reconstruction task, the main flow is: + * - ListBlock from all healthy replicas + * - calculate effective block group len for all blocks + * - for each block + * - build a ReconstructInputStream to read healthy chunks + * - build a ECBlockOutputStream to write out decoded chunks + * - for each stripe + * - use ReconstructInputStream.readStripe to decode missing chunks + * - use ECBlockOutputStream.write to write decoded chunks to TargetDNs + * - PutBlock + * - CloseContainer + */ +public class ECReconstructionCoordinator implements Closeable { + + static final Logger LOG = + LoggerFactory.getLogger(ECReconstructionCoordinator.class); + + private static final int EC_RECONSTRUCT_STRIPE_READ_POOL_MIN_SIZE = 3; + + private final ECContainerOperationClient containerOperationClient; + + private final ConfigurationSource config; + + private final ByteBufferPool byteBufferPool; + + private ExecutorService ecReconstructExecutor; + + private BlockInputStreamFactory blockInputStreamFactory; + + public ECReconstructionCoordinator(ECContainerOperationClient containerClient, + ConfigurationSource conf, ByteBufferPool byteBufferPool, + ExecutorService reconstructExecutor, + BlockInputStreamFactory streamFactory) { + this.containerOperationClient = containerClient; + this.config = conf; + this.byteBufferPool = byteBufferPool; + this.blockInputStreamFactory = streamFactory; + this.ecReconstructExecutor = reconstructExecutor; + } + + public ECReconstructionCoordinator(ConfigurationSource conf, + CertificateClient certificateClient) throws IOException { + this(new ECContainerOperationClient(conf, certificateClient), conf, + new ElasticByteBufferPool(), null, null); + this.ecReconstructExecutor = + new ThreadPoolExecutor(EC_RECONSTRUCT_STRIPE_READ_POOL_MIN_SIZE, + config.getObject(OzoneClientConfig.class) + .getEcReconstructStripeReadPoolLimit(), 60, TimeUnit.SECONDS, + new SynchronousQueue<>(), new ThreadFactoryBuilder() + .setNameFormat("ec-reconstruct-reader-TID-%d").build(), + new ThreadPoolExecutor.CallerRunsPolicy()); + this.blockInputStreamFactory = BlockInputStreamFactoryImpl + .getInstance(byteBufferPool, () -> ecReconstructExecutor); + } + + public void reconstructECContainerGroup(long containerID, + ECReplicationConfig repConfig, + SortedMap<Integer, DatanodeDetails> sourceNodeMap, + SortedMap<Integer, DatanodeDetails> targetNodeMap) throws IOException { + + Pipeline pipeline = rebuildInputPipeline(repConfig, sourceNodeMap); + + SortedMap<Long, BlockData[]> blockDataMap = + getBlockDataMap(containerID, repConfig, sourceNodeMap); + + SortedMap<Long, BlockLocationInfo> blockLocationInfoMap = + calcBlockLocationInfoMap(containerID, blockDataMap, pipeline); + + // 1. create target recovering containers. + Set<Map.Entry<Integer, DatanodeDetails>> targetIndexDns = + targetNodeMap.entrySet(); + Iterator<Map.Entry<Integer, DatanodeDetails>> iterator = + targetIndexDns.iterator(); + while (iterator.hasNext()) { + Map.Entry<Integer, DatanodeDetails> indexDnPair = iterator.next(); + this.containerOperationClient + .createRecoveringContainer(containerID, indexDnPair.getValue(), + repConfig, null, indexDnPair.getKey()); + } + + // 2. Reconstruct and transfer to targets + for (BlockLocationInfo blockLocationInfo : blockLocationInfoMap.values()) { + reconstructECBlockGroup(blockLocationInfo, repConfig, targetNodeMap); + } + + // 3. Close containers + iterator = targetIndexDns.iterator(); + while (iterator.hasNext()) { + DatanodeDetails dn = iterator.next().getValue(); + this.containerOperationClient + .closeContainer(containerID, dn, repConfig, null); + } + + } + + void reconstructECBlockGroup(BlockLocationInfo blockLocationInfo, + ECReplicationConfig repConfig, + SortedMap<Integer, DatanodeDetails> targetMap) + throws IOException { + long safeBlockGroupLength = blockLocationInfo.getLength(); + List<Integer> missingContainerIndexes = + targetMap.keySet().stream().collect(Collectors.toList()); + + // calculate the real missing block indexes + int dataLocs = ECBlockInputStreamProxy + .expectedDataLocations(repConfig, safeBlockGroupLength); + List<Integer> toReconstructIndexes = new ArrayList<>(); + for (int i = 0; i < missingContainerIndexes.size(); i++) { + Integer index = missingContainerIndexes.get(i); + if (index <= dataLocs || index > repConfig.getData()) { + toReconstructIndexes.add(index); + } + // else padded indexes. + } + + // Looks like we don't need to reconstruct any missing blocks in this block + // group. The reason for this should be block group had only padding blocks + // in the missing locations. + if (toReconstructIndexes.size() == 0) { + if (LOG.isDebugEnabled()) { + LOG.debug("Skipping the reconstruction for the block: " + + blockLocationInfo.getBlockID() + ". In the missing locations: " + + missingContainerIndexes + + ", this block group has only padded blocks."); + } + return; + } + + try (ECBlockReconstructedStripeInputStream sis + = new ECBlockReconstructedStripeInputStream( + repConfig, blockLocationInfo, true, + this.containerOperationClient.getXceiverClientManager(), null, + this.blockInputStreamFactory, byteBufferPool, + this.ecReconstructExecutor)) { + + ECBlockOutputStream[] targetBlockStreams = + new ECBlockOutputStream[toReconstructIndexes.size()]; + ByteBuffer[] bufs = new ByteBuffer[toReconstructIndexes.size()]; + for (int i = 0; i < toReconstructIndexes.size(); i++) { + OzoneClientConfig configuration = new OzoneClientConfig(); + // TODO: Let's avoid unnecessary bufferPool creation for + BufferPool bufferPool = + new BufferPool(configuration.getStreamBufferSize(), + (int) (configuration.getStreamBufferMaxSize() / configuration + .getStreamBufferSize()), + ByteStringConversion.createByteBufferConversion(false)); + targetBlockStreams[i] = + new ECBlockOutputStream(blockLocationInfo.getBlockID(), + this.containerOperationClient.getXceiverClientManager(), + Pipeline.newBuilder().setId(PipelineID.valueOf( + targetMap.get(toReconstructIndexes.get(i)).getUuid())) + .setReplicationConfig(repConfig).setNodes(ImmutableList + .of(targetMap.get(toReconstructIndexes.get(i)))) + .setState(Pipeline.PipelineState.CLOSED).build(), Review Comment: done ########## hadoop-hdds/client/src/main/java/org/apache/hadoop/hdds/scm/storage/ECBlockOutputStream.java: ########## @@ -80,6 +80,10 @@ public void write(byte[] b, int off, int len) throws IOException { writeChunkToContainer(ChunkBuffer.wrap(ByteBuffer.wrap(b, off, len))); } + public void write(ByteBuffer buff) throws IOException { + this.currentChunkRspFuture = writeChunkToContainer(ChunkBuffer.wrap(buff)); Review Comment: sure -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
