gianm commented on code in PR #17917: URL: https://github.com/apache/druid/pull/17917#discussion_r2046116432
########## indexing-service/src/main/java/org/apache/druid/indexing/seekablestream/SeekableStreamAppenderatorConfig.java: ########## @@ -0,0 +1,282 @@ +/* + * 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.druid.indexing.seekablestream; + +import org.apache.druid.indexer.partitions.PartitionsSpec; +import org.apache.druid.java.util.common.logger.Logger; +import org.apache.druid.query.DruidProcessingConfig; +import org.apache.druid.segment.IndexSpec; +import org.apache.druid.segment.incremental.AppendableIndexSpec; +import org.apache.druid.segment.indexing.TuningConfig; +import org.apache.druid.segment.realtime.appenderator.AppenderatorConfig; +import org.apache.druid.segment.writeout.SegmentWriteOutMediumFactory; +import org.apache.druid.utils.JvmUtils; +import org.apache.druid.utils.RuntimeInfo; +import org.joda.time.Period; + +import javax.annotation.Nullable; +import java.io.File; +import java.util.Objects; + +public class SeekableStreamAppenderatorConfig implements AppenderatorConfig +{ + private static final Logger log = new Logger(SeekableStreamAppenderatorConfig.class); + + /** + * Rough estimate of the on-heap overhead of reading a column. Determined empirically through heap dumps. + */ + public static final int APPENDERATOR_MERGE_ROUGH_HEAP_MEMORY_PER_COLUMN = 3_000; + + /** + * Rough estimate of the off-heap overhead of reading a column. This is 1.5x the size of a decompression buffer. + * The 1.5x allows for some headroom for unaccounted-for memory usage. The decompression buffer size is about right + * when a compressing {@link IndexSpec} is in play. It is an overestimate when compression is disabled for + * intermediate persists, but this isn't currently the default behavior, so we're wanting to be conservative here. + * + * See also {@link AppenderatorConfig#getIndexSpecForIntermediatePersists()} for how compression can be disabled + * for intermediate persists. + */ + public static final int APPENDERATOR_MERGE_ROUGH_DIRECT_MEMORY_PER_COLUMN = (int) ((1 << 16) * 1.5); + + private final SeekableStreamIndexTaskTuningConfig tuningConfig; + private final int maxColumnsToMerge; + + private SeekableStreamAppenderatorConfig(SeekableStreamIndexTaskTuningConfig tuningConfig, int maxColumnsToMerge) + { + this.tuningConfig = tuningConfig; + this.maxColumnsToMerge = maxColumnsToMerge; + } + + public static SeekableStreamAppenderatorConfig fromTuningConfig( + final SeekableStreamIndexTaskTuningConfig tuningConfig, + @Nullable final DruidProcessingConfig processingConfig + ) + { + if (processingConfig == null) { + return new SeekableStreamAppenderatorConfig(tuningConfig, tuningConfig.getMaxColumnsToMerge()); + } else { + final int maxColumnsToMerge; + if (tuningConfig.getMaxColumnsToMerge() == SeekableStreamIndexTaskTuningConfig.DEFAULT_MAX_COLUMNS_TO_MERGE) { + maxColumnsToMerge = calculateDefaultMaxColumnsToMerge( + JvmUtils.getRuntimeInfo(), + processingConfig, + tuningConfig + ); + } else { + maxColumnsToMerge = tuningConfig.getMaxColumnsToMerge(); + } + + return new SeekableStreamAppenderatorConfig(tuningConfig, maxColumnsToMerge); + } + } + + @Override + public boolean isReportParseExceptions() + { + return tuningConfig.isReportParseExceptions(); + } + + @Override + public int getMaxPendingPersists() + { + return tuningConfig.getMaxPendingPersists(); + } + + @Override + public boolean isSkipBytesInMemoryOverheadCheck() + { + return tuningConfig.isSkipBytesInMemoryOverheadCheck(); + } + + @Override + public Period getIntermediatePersistPeriod() + { + return tuningConfig.getIntermediatePersistPeriod(); + } + + @Override + public File getBasePersistDirectory() + { + return tuningConfig.getBasePersistDirectory(); + } + + @Override + public AppenderatorConfig withBasePersistDirectory(File basePersistDirectory) + { + return new SeekableStreamAppenderatorConfig( + tuningConfig.withBasePersistDirectory(basePersistDirectory), + maxColumnsToMerge + ); + } + + @Override + public SegmentWriteOutMediumFactory getSegmentWriteOutMediumFactory() + { + return tuningConfig.getSegmentWriteOutMediumFactory(); + } + + @Override + public AppendableIndexSpec getAppendableIndexSpec() + { + return tuningConfig.getAppendableIndexSpec(); + } + + @Override + public int getMaxRowsInMemory() + { + return tuningConfig.getMaxRowsInMemory(); + } + + @Override + public long getMaxBytesInMemory() + { + return tuningConfig.getMaxBytesInMemory(); + } + + @Override + public PartitionsSpec getPartitionsSpec() + { + return tuningConfig.getPartitionsSpec(); + } + + @Override + public IndexSpec getIndexSpec() + { + return tuningConfig.getIndexSpec(); + } + + @Override + public IndexSpec getIndexSpecForIntermediatePersists() + { + return tuningConfig.getIndexSpecForIntermediatePersists(); + } + + @Override + public int getNumPersistThreads() + { + return tuningConfig.getNumPersistThreads(); + } + + @Override + public Integer getMaxRowsPerSegment() + { + return tuningConfig.getMaxRowsPerSegment(); + } + + @Override + public Long getMaxTotalRows() + { + return tuningConfig.getMaxTotalRows(); + } + + @Override + public int getMaxColumnsToMerge() + { + return maxColumnsToMerge; + } + + @Override + public boolean equals(Object o) + { + if (o == null || getClass() != o.getClass()) { + return false; + } + SeekableStreamAppenderatorConfig that = (SeekableStreamAppenderatorConfig) o; + return maxColumnsToMerge == that.maxColumnsToMerge && Objects.equals(tuningConfig, that.tuningConfig); + } + + @Override + public int hashCode() + { + return Objects.hash(tuningConfig, maxColumnsToMerge); + } + + /** + * Calculates the value to use for "maxColumnsToMerge" when it has been set to + * {@link SeekableStreamIndexTaskTuningConfig#DEFAULT_MAX_COLUMNS_TO_MERGE}. + */ + static int calculateDefaultMaxColumnsToMerge( + final RuntimeInfo runtimeInfo, + final DruidProcessingConfig processingConfig, + final TuningConfig tuningConfig + ) + { + final int maxColumnsToMergeBasedOnHeapMemory = getMaxColumnsToMergeBasedOnHeapMemory(tuningConfig); + final int maxColumnsToMergeBasedOnDirectMemory = + getMaxColumnsToMergeBasedOnDirectMemory(runtimeInfo, processingConfig); + + return Math.min(maxColumnsToMergeBasedOnHeapMemory, maxColumnsToMergeBasedOnDirectMemory); + } + + /** + * Calculates the value to use for "maxColumnsToMerge" based purely on heap memory. + */ + static int getMaxColumnsToMergeBasedOnHeapMemory(TuningConfig tuningConfig) + { + final long heapMemoryForMerging; + + if (tuningConfig.getMaxBytesInMemory() >= 0) { + // maxBytesInMemory is active and represents the amount of heap memory used for indexing. + // Use an equal amount of heap memory for merging. + heapMemoryForMerging = tuningConfig.getMaxBytesInMemoryOrDefault(); + } else { + // maxBytesInMemory was set to unlimited by the user; use what that _would_ be if it was set to automatic. + heapMemoryForMerging = tuningConfig.getAppendableIndexSpec().getDefaultMaxBytesInMemory(); + } + + return (int) Math.min(heapMemoryForMerging / APPENDERATOR_MERGE_ROUGH_HEAP_MEMORY_PER_COLUMN, Integer.MAX_VALUE); + } + + /** + * Calculates the value to use for "maxColumnsToMerge" based purely on direct memory. + */ + static int getMaxColumnsToMergeBasedOnDirectMemory( + final RuntimeInfo runtimeInfo, + final DruidProcessingConfig processingConfig + ) + { + final long directMemoryForProcessing = + (long) processingConfig.intermediateComputeSizeBytes() * + (processingConfig.getNumThreads() + processingConfig.getNumMergeBuffers()); Review Comment: Renamed the variable to `directMemoryForQueries` and added this comment: > // Query processing requires one processing buffer per thread, plus all the merge buffers. Hopefully that is more clear. The idea is this is meant to track the size of both the processing pool and merge pool added together. -- 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]
