aokolnychyi commented on a change in pull request #2591: URL: https://github.com/apache/iceberg/pull/2591#discussion_r635500520
########## File path: spark3/src/main/java/org/apache/iceberg/spark/actions/Spark3BinPackStrategy.java ########## @@ -0,0 +1,88 @@ +/* + * 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.iceberg.spark.actions; + +import java.util.List; +import java.util.Set; +import org.apache.iceberg.DataFile; +import org.apache.iceberg.FileScanTask; +import org.apache.iceberg.Table; +import org.apache.iceberg.actions.BinPackStrategy; +import org.apache.iceberg.relocated.com.google.common.collect.ImmutableSet; +import org.apache.iceberg.spark.FileRewriteCoordinator; +import org.apache.iceberg.spark.FileScanTaskSetManager; +import org.apache.iceberg.spark.SparkReadOptions; +import org.apache.iceberg.spark.SparkWriteOptions; +import org.apache.spark.sql.Dataset; +import org.apache.spark.sql.Row; +import org.apache.spark.sql.SparkSession; +import org.apache.spark.sql.internal.SQLConf; + +public class Spark3BinPackStrategy extends BinPackStrategy { + private final Table table; + private final SparkSession spark; + private final FileScanTaskSetManager manager = FileScanTaskSetManager.get(); + private final FileRewriteCoordinator rewriteCoordinator = FileRewriteCoordinator.get(); + + public Spark3BinPackStrategy(Table table, SparkSession spark) { + this.table = table; + this.spark = spark; + } + + @Override + public Table table() { + return table; + } + + @Override + public Set<DataFile> rewriteFiles(String groupID, List<FileScanTask> filesToRewrite) { + manager.stageTasks(table, groupID, filesToRewrite); + + // Disable Adaptive Query Execution as this may change the output partitioning of our write + SparkSession cloneSession = spark.cloneSession(); + cloneSession.conf().set(SQLConf.ADAPTIVE_EXECUTION_ENABLED().key(), false); + + Dataset<Row> scanDF = cloneSession.read().format("iceberg") + .option(SparkReadOptions.FILE_SCAN_TASK_SET_ID, groupID) + .option(SparkReadOptions.SPLIT_SIZE, Long.toString(targetFileSize())) + .option(SparkReadOptions.FILE_OPEN_COST, "0") + .load(table.name()); + + // write the packed data into new files where each split becomes a new file + + try { + scanDF.write() Review comment: There is two situations we have to be careful with. First, the result file size may be a little bit bigger than our target file size. We don't want to cut another file with just a couple of MBs. It is better to write a slightly bigger file, especially because we have the max file size threshold. Second, I have seen a number of use cases when the file size after compaction is substantially larger than what we anticipated. I suspect it is related to not being able to apply a specific encoding technique as the size of a row group grows. There were use cases when there was more than 100% mismatch. I think these two use cases can be solved by limiting our file size in the compaction as let's say 0.8 (or any other value) of the max file size. This can be done by passing a dedicated write option. ########## File path: spark3/src/main/java/org/apache/iceberg/spark/actions/Spark3BinPackStrategy.java ########## @@ -0,0 +1,88 @@ +/* + * 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.iceberg.spark.actions; + +import java.util.List; +import java.util.Set; +import org.apache.iceberg.DataFile; +import org.apache.iceberg.FileScanTask; +import org.apache.iceberg.Table; +import org.apache.iceberg.actions.BinPackStrategy; +import org.apache.iceberg.relocated.com.google.common.collect.ImmutableSet; +import org.apache.iceberg.spark.FileRewriteCoordinator; +import org.apache.iceberg.spark.FileScanTaskSetManager; +import org.apache.iceberg.spark.SparkReadOptions; +import org.apache.iceberg.spark.SparkWriteOptions; +import org.apache.spark.sql.Dataset; +import org.apache.spark.sql.Row; +import org.apache.spark.sql.SparkSession; +import org.apache.spark.sql.internal.SQLConf; + +public class Spark3BinPackStrategy extends BinPackStrategy { + private final Table table; + private final SparkSession spark; + private final FileScanTaskSetManager manager = FileScanTaskSetManager.get(); + private final FileRewriteCoordinator rewriteCoordinator = FileRewriteCoordinator.get(); + + public Spark3BinPackStrategy(Table table, SparkSession spark) { + this.table = table; + this.spark = spark; + } + + @Override + public Table table() { + return table; + } + + @Override + public Set<DataFile> rewriteFiles(String groupID, List<FileScanTask> filesToRewrite) { + manager.stageTasks(table, groupID, filesToRewrite); + + // Disable Adaptive Query Execution as this may change the output partitioning of our write + SparkSession cloneSession = spark.cloneSession(); + cloneSession.conf().set(SQLConf.ADAPTIVE_EXECUTION_ENABLED().key(), false); + + Dataset<Row> scanDF = cloneSession.read().format("iceberg") + .option(SparkReadOptions.FILE_SCAN_TASK_SET_ID, groupID) + .option(SparkReadOptions.SPLIT_SIZE, Long.toString(targetFileSize())) + .option(SparkReadOptions.FILE_OPEN_COST, "0") + .load(table.name()); + + // write the packed data into new files where each split becomes a new file + + try { + scanDF.write() Review comment: There are two situations we have to be careful with. First, the result file size may be a little bit bigger than our target file size. We don't want to cut another file with just a couple of MBs. It is better to write a slightly bigger file, especially because we have the max file size threshold. Second, I have seen a number of use cases when the file size after compaction is substantially larger than what we anticipated. I suspect it is related to not being able to apply a specific encoding technique as the size of a row group grows. There were use cases when there was more than 100% mismatch. I think these two use cases can be solved by limiting our file size in the compaction as let's say 0.8 (or any other value) of the max file size. This can be done by passing a dedicated write option. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
