sjwiesman commented on a change in pull request #8861: [FLINK-12963][state-processor-api] Add savepoint writer for bootstrapping new savepoints URL: https://github.com/apache/flink/pull/8861#discussion_r297760624
########## File path: flink-libraries/flink-state-processing-api/src/main/java/org/apache/flink/state/api/BootstrapTransformation.java ########## @@ -0,0 +1,129 @@ +/* + * 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.state.api; + +import org.apache.flink.annotation.PublicEvolving; +import org.apache.flink.api.common.typeinfo.TypeInformation; +import org.apache.flink.api.common.typeutils.TypeSerializer; +import org.apache.flink.api.java.DataSet; +import org.apache.flink.api.java.functions.KeySelector; +import org.apache.flink.api.java.operators.MapPartitionOperator; +import org.apache.flink.core.fs.Path; +import org.apache.flink.runtime.state.StateBackend; +import org.apache.flink.state.api.output.BoundedOneInputStreamTaskRunner; +import org.apache.flink.state.api.output.TaggedOperatorSubtaskState; +import org.apache.flink.state.api.output.operators.BroadcastStateBootstrapOperator; +import org.apache.flink.state.api.output.partitioner.HashSelector; +import org.apache.flink.state.api.output.partitioner.KeyGroupRangePartitioner; +import org.apache.flink.state.api.runtime.BoundedStreamConfig; +import org.apache.flink.state.api.runtime.OperatorIDGenerator; +import org.apache.flink.state.api.runtime.metadata.SavepointMetadata; +import org.apache.flink.streaming.api.graph.StreamConfig; +import org.apache.flink.streaming.api.operators.StreamOperator; + +import javax.annotation.Nonnull; +import javax.annotation.Nullable; + +/** + * Bootstrapped data that can be written into a {@code Savepoint}. + * @param <T> The input type of the transformation. + */ +@PublicEvolving +@SuppressWarnings("WeakerAccess") +public class BootstrapTransformation<T> { + private final DataSet<T> dataSet; + + private final SavepointWriterOperatorFactory factory; + + @Nullable + private final HashSelector<T> keySelector; + + @Nullable + private final TypeInformation<?> keyType; + + BootstrapTransformation( + DataSet<T> dataSet, + SavepointWriterOperatorFactory factory) { + this.dataSet = dataSet; + this.factory = factory; + this.keySelector = null; + this.keyType = null; + } + + <K> BootstrapTransformation( + DataSet<T> dataSet, + SavepointWriterOperatorFactory factory, + @Nonnull KeySelector<T, K> keySelector, + @Nonnull TypeInformation<K> keyType) { + this.dataSet = dataSet; + this.factory = factory; + this.keySelector = new HashSelector<>(keySelector); + this.keyType = keyType; + } + + DataSet<TaggedOperatorSubtaskState> getOperatorSubtaskStates( + String uid, + StateBackend stateBackend, + SavepointMetadata metadata, + Path savepointPath) { + DataSet<T> input = dataSet; + if (keySelector != null) { + input = dataSet.partitionCustom(new KeyGroupRangePartitioner(metadata.maxParallelism()), keySelector); + } + + final StreamConfig config; + if (keyType == null) { + config = new BoundedStreamConfig(); + } else { + TypeSerializer<?> keySerializer = keyType.createSerializer(dataSet.getExecutionEnvironment().getConfig()); + config = new BoundedStreamConfig(keySerializer, keySelector); + } + + StreamOperator<TaggedOperatorSubtaskState> operator = factory.getOperator( + System.currentTimeMillis(), + savepointPath); + + operator = dataSet.clean(operator); + config.setStreamOperator(operator); + + config.setOperatorName(uid); + config.setOperatorID(OperatorIDGenerator.fromUid(uid)); + config.setStateBackend(stateBackend); + + BoundedOneInputStreamTaskRunner<T> operatorRunner = new BoundedOneInputStreamTaskRunner<>( + config, + metadata + ); + + MapPartitionOperator<T, TaggedOperatorSubtaskState> subtaskStates = input + .mapPartition(operatorRunner) + .name(uid); + + if (operator instanceof BroadcastStateBootstrapOperator) { + subtaskStates = subtaskStates.setParallelism(1); Review comment: Because the DataSet api does not support broadcast partitioning. This is the closest approximation; if the parallelism is one than all elements will be shipped to the same task slot. In practice it's fine because we only need a single copy of the broadcast state which can then be restored at any parallelism. ---------------------------------------------------------------- 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: [email protected] With regards, Apache Git Services
