pvary commented on code in PR #11497: URL: https://github.com/apache/iceberg/pull/11497#discussion_r1870222639
########## flink/v1.20/flink/src/main/java/org/apache/iceberg/flink/maintenance/operator/DataFileRewriteCommitter.java: ########## @@ -0,0 +1,304 @@ +/* + * 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.flink.maintenance.operator; + +import java.io.IOException; +import java.util.Iterator; +import java.util.Set; +import org.apache.flink.annotation.Internal; +import org.apache.flink.api.common.state.ListState; +import org.apache.flink.api.common.state.ListStateDescriptor; +import org.apache.flink.api.common.typeinfo.BasicTypeInfo; +import org.apache.flink.api.common.typeinfo.TypeInformation; +import org.apache.flink.metrics.Counter; +import org.apache.flink.metrics.MetricGroup; +import org.apache.flink.runtime.state.StateInitializationContext; +import org.apache.flink.runtime.state.StateSnapshotContext; +import org.apache.flink.streaming.api.operators.AbstractStreamOperator; +import org.apache.flink.streaming.api.operators.OneInputStreamOperator; +import org.apache.flink.streaming.api.watermark.Watermark; +import org.apache.flink.streaming.runtime.streamrecord.StreamRecord; +import org.apache.iceberg.DataFile; +import org.apache.iceberg.FileContent; +import org.apache.iceberg.Table; +import org.apache.iceberg.actions.RewriteDataFilesCommitManager; +import org.apache.iceberg.actions.RewriteDataFilesCommitManager.CommitService; +import org.apache.iceberg.actions.RewriteFileGroup; +import org.apache.iceberg.flink.TableLoader; +import org.apache.iceberg.flink.maintenance.api.Trigger; +import org.apache.iceberg.relocated.com.google.common.base.Preconditions; +import org.apache.iceberg.relocated.com.google.common.collect.Sets; +import org.slf4j.Logger; +import org.slf4j.LoggerFactory; + +/** + * Commits the compaction changes using {@link RewriteDataFilesCommitManager}. The input is a {@link + * DataFileRewriteExecutor.ExecutedGroup}. Only {@link Watermark} is emitted which is chained to + * {@link TaskResultAggregator} input 1. + */ +@Internal +public class DataFileRewriteCommitter extends AbstractStreamOperator<Trigger> + implements OneInputStreamOperator<DataFileRewriteExecutor.ExecutedGroup, Trigger> { + private static final Logger LOG = LoggerFactory.getLogger(DataFileRewriteCommitter.class); + + private final String tableName; + private final String taskName; + private final int taskIndex; + private final TableLoader tableLoader; + + private transient Table table; + private transient Long startingSnapshotId; + private transient Set<RewriteFileGroup> inProgress; + private transient ListState<Long> startingSnapshotIdState; + private transient ListState<RewriteFileGroup> inProgressState; + private transient CommitService commitService; + private transient Counter errorCounter; + private transient Counter addedDataFileNumCounter; + private transient Counter addedDataFileSizeCounter; + private transient Counter removedDataFileNumCounter; + private transient Counter removedDataFileSizeCounter; + + public DataFileRewriteCommitter( + String tableName, String taskName, int taskIndex, TableLoader tableLoader) { + Preconditions.checkNotNull(tableName, "Table name should no be null"); + Preconditions.checkNotNull(taskName, "Task name should no be null"); + Preconditions.checkNotNull(tableLoader, "Table loader should no be null"); + + this.tableName = tableName; + this.taskName = taskName; + this.taskIndex = taskIndex; + this.tableLoader = tableLoader; + } + + @Override + public void open() throws Exception { + MetricGroup taskMetricGroup = + TableMaintenanceMetrics.groupFor(getRuntimeContext(), tableName, taskName, taskIndex); + this.errorCounter = taskMetricGroup.counter(TableMaintenanceMetrics.ERROR_COUNTER); + this.addedDataFileNumCounter = + taskMetricGroup.counter(TableMaintenanceMetrics.ADDED_DATA_FILE_NUM_METRIC); + this.addedDataFileSizeCounter = + taskMetricGroup.counter(TableMaintenanceMetrics.ADDED_DATA_FILE_SIZE_METRIC); + this.removedDataFileNumCounter = + taskMetricGroup.counter(TableMaintenanceMetrics.REMOVED_DATA_FILE_NUM_METRIC); + this.removedDataFileSizeCounter = + taskMetricGroup.counter(TableMaintenanceMetrics.REMOVED_DATA_FILE_SIZE_METRIC); + + tableLoader.open(); + this.table = tableLoader.loadTable(); + } + + @Override + public void initializeState(StateInitializationContext context) throws Exception { + super.initializeState(context); + this.startingSnapshotIdState = + context + .getOperatorStateStore() + .getListState( + new ListStateDescriptor<>( + "data-file-updater-snapshot-id", BasicTypeInfo.LONG_TYPE_INFO)); + this.inProgressState = + context + .getOperatorStateStore() + .getListState( + new ListStateDescriptor<>( + "data-file-updater-in-progress", TypeInformation.of(RewriteFileGroup.class))); + + this.startingSnapshotId = null; + Iterable<Long> snapshotIdIterable = startingSnapshotIdState.get(); + if (snapshotIdIterable != null) { + Iterator<Long> snapshotIdIterator = snapshotIdIterable.iterator(); + if (snapshotIdIterator.hasNext()) { + this.startingSnapshotId = snapshotIdIterator.next(); + } + } + + this.inProgress = Sets.newHashSet(); + Iterable<RewriteFileGroup> inProgressIterable = inProgressState.get(); + if (inProgressIterable != null) { + for (RewriteFileGroup group : inProgressIterable) { + inProgress.add(group); + } + } + } + + @Override + public void snapshotState(StateSnapshotContext context) throws Exception { + super.snapshotState(context); + + startingSnapshotIdState.clear(); + if (startingSnapshotId != null) { + startingSnapshotIdState.add(startingSnapshotId); + } + + inProgressState.clear(); + for (RewriteFileGroup group : inProgress) { + inProgressState.add(group); + } + } + + @Override + public void processElement(StreamRecord<DataFileRewriteExecutor.ExecutedGroup> streamRecord) { + DataFileRewriteExecutor.ExecutedGroup executedGroup = streamRecord.getValue(); + try { + if (commitService == null) { + this.commitService = createCommitService(executedGroup, streamRecord.getTimestamp()); + this.startingSnapshotId = executedGroup.snapshotId(); + } + + commitService.offer(executedGroup.group()); + inProgress.add(executedGroup.group()); + } catch (Exception e) { + LOG.info( + "Exception processing {} for table {} with {}[{}] at {}", + executedGroup, + tableName, + taskName, + taskIndex, + streamRecord.getTimestamp(), + e); + output.collect(TaskResultAggregator.ERROR_STREAM, new StreamRecord<>(e)); + errorCounter.inc(); + } + } + + @Override + public void processWatermark(Watermark mark) throws Exception { + try { + if (commitService == null && !inProgress.isEmpty()) { + this.commitService = createCommitService(null, mark.getTimestamp()); + } + + if (commitService != null) { + commitService.close(); + } + + table.refresh(); + LOG.info( + "Successfully completed data file compaction to {} for table {} with {}[{}] at {}", + table.currentSnapshot().snapshotId(), + tableName, + taskName, + taskIndex, + mark.getTimestamp()); + } catch (Exception e) { + LOG.info( + "Exception closing commit service for table {} with {}[{}] at {}", + tableName, + taskName, + taskIndex, + mark.getTimestamp(), + e); + output.collect(TaskResultAggregator.ERROR_STREAM, new StreamRecord<>(e)); + errorCounter.inc(); + } + + // Cleanup + commitService = null; + startingSnapshotId = null; + inProgress.clear(); + + super.processWatermark(mark); + } + + @Override + public void close() throws IOException { + if (commitService != null) { + commitService.close(); Review Comment: What we need to do: Basically we would need to check every group if it was committed or not. This would mean to keep an up-to-date cache for the already committed groupIds (timestamp + globalId). This also could become "corrupt" if the grouping algorithm or configuration changes. What we gain: without this we try and validation would prevent duplicate commits. This would happen only during restarts where unaligned checkpointing is used. I'm starting to get convinced to basically just remove the state from the committer... When "close" is called on the `CommitService` then every pending group is committed anyways. If we lose some groups, then we lose them... not a big deal. On job restart or on the next compaction they will be regenerated anyways. Seems like it is less error prone if we just forget about the pending commits and try to commit every group which we receive. -- 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: issues-unsubscr...@iceberg.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@iceberg.apache.org For additional commands, e-mail: issues-h...@iceberg.apache.org
