This is an automated email from the ASF dual-hosted git repository. haibow pushed a commit to branch release-0.4.0-rc-2 in repository https://gitbox.apache.org/repos/asf/incubator-pinot.git
commit fe3b98ea8d2283d4b67b0bdb8622cbf4490f2cc1 Author: Sidd <siddharthteo...@gmail.com> AuthorDate: Sat May 30 19:21:23 2020 -0700 Two changes: (#5470) (1) PR https://github.com/apache/incubator-pinot/pull/5256 added support for deriving num docs per chunk for var byte raw index create from column length. This was specifically done as part of supporting text blobs. For use cases that don't want this feature and are high QPS, see a negative impact since size of chunk increases (earlier value of numDocsPerChunk was hardcoded to 1000) and based on the access pattern we might end up uncompressing a bigger chunk to get values for a set of docIds. We have made this change configurable. So the default behaviour is same as old (1000 docs per chunk) (2) PR https://github.com/apache/incubator-pinot/pull/4791 added support for noDict for STRING/BYTES in consuming segments. There is a particular impact of this change on the use cases that have set noDict on their STRING dimension columns for other performance reasons and also want metricsAggregation. These use cases don't get to aggregateMetrics because the new implementation was able to honor their table config setting of noDict on STRING/BYTES. Without metrics aggregation, memory pressure increases. So to continue aggregating metrics for such cases, we will create dictionary even if the column is part of noDictionary set from table config. Co-authored-by: Siddharth Teotia <steo...@steotia-mn1.linkedin.biz> --- .../generator/SegmentGeneratorConfig.java | 15 +++++++++++ .../indexsegment/mutable/MutableSegmentImpl.java | 29 +++++++++++++++++++--- .../creator/impl/SegmentColumnarIndexCreator.java | 22 ++++++++++++++-- .../fwd/SingleValueVarByteRawIndexCreator.java | 10 +++++++- .../defaultcolumn/BaseDefaultColumnHandler.java | 3 ++- .../apache/pinot/spi/config/table/FieldConfig.java | 1 + 6 files changed, 72 insertions(+), 8 deletions(-) diff --git a/pinot-core/src/main/java/org/apache/pinot/core/indexsegment/generator/SegmentGeneratorConfig.java b/pinot-core/src/main/java/org/apache/pinot/core/indexsegment/generator/SegmentGeneratorConfig.java index 59531fe..9af9c16 100644 --- a/pinot-core/src/main/java/org/apache/pinot/core/indexsegment/generator/SegmentGeneratorConfig.java +++ b/pinot-core/src/main/java/org/apache/pinot/core/indexsegment/generator/SegmentGeneratorConfig.java @@ -102,6 +102,9 @@ public class SegmentGeneratorConfig { private boolean _skipTimeValueCheck = false; private boolean _nullHandlingEnabled = false; + // constructed from FieldConfig + private Map<String, Map<String, String>> _columnProperties = new HashMap<>(); + @Deprecated public SegmentGeneratorConfig() { } @@ -174,12 +177,24 @@ public class SegmentGeneratorConfig { _invertedIndexCreationColumns = indexingConfig.getInvertedIndexColumns(); } + List<FieldConfig> fieldConfigList = tableConfig.getFieldConfigList(); + if (fieldConfigList != null) { + for (FieldConfig fieldConfig : fieldConfigList) { + _columnProperties.put(fieldConfig.getName(), fieldConfig.getProperties()); + } + } + extractTextIndexColumnsFromTableConfig(tableConfig); _nullHandlingEnabled = indexingConfig.isNullHandlingEnabled(); } } + @Nonnull + public Map<String, Map<String, String>> getColumnProperties() { + return _columnProperties; + } + /** * Set time column details using the given time column */ diff --git a/pinot-core/src/main/java/org/apache/pinot/core/indexsegment/mutable/MutableSegmentImpl.java b/pinot-core/src/main/java/org/apache/pinot/core/indexsegment/mutable/MutableSegmentImpl.java index 07e5ec9..0e24664 100644 --- a/pinot-core/src/main/java/org/apache/pinot/core/indexsegment/mutable/MutableSegmentImpl.java +++ b/pinot-core/src/main/java/org/apache/pinot/core/indexsegment/mutable/MutableSegmentImpl.java @@ -219,7 +219,7 @@ public class MutableSegmentImpl implements MutableSegment { FieldSpec.DataType dataType = fieldSpec.getDataType(); boolean isFixedWidthColumn = dataType.isFixedWidth(); int forwardIndexColumnSize = -1; - if (isNoDictionaryColumn(noDictionaryColumns, invertedIndexColumns, textIndexColumns, fieldSpec, column)) { + if (isNoDictionaryColumn(noDictionaryColumns, invertedIndexColumns, fieldSpec, column)) { // no dictionary // each forward index entry will be equal to size of data for that row // For INT, LONG, FLOAT, DOUBLE it is equal to the number of fixed bytes used to store the value, @@ -329,9 +329,30 @@ public class MutableSegmentImpl implements MutableSegment { * @return true if column is no-dictionary, false if dictionary encoded */ private boolean isNoDictionaryColumn(Set<String> noDictionaryColumns, Set<String> invertedIndexColumns, - Set<String> textIndexColumns, FieldSpec fieldSpec, String column) { - return textIndexColumns.contains(column) || (noDictionaryColumns.contains(column) && fieldSpec.isSingleValueField() - && !invertedIndexColumns.contains(column)); + FieldSpec fieldSpec, String column) { + FieldSpec.DataType dataType = fieldSpec.getDataType(); + if (noDictionaryColumns.contains(column)) { + // Earlier we didn't support noDict in consuming segments for STRING and BYTES columns. + // So even if the user had the column in noDictionaryColumns set in table config, we still + // created dictionary in consuming segments. + // Later on we added this support. There is a particular impact of this change on the use cases + // that have set noDict on their STRING dimension columns for other performance + // reasons and also want metricsAggregation. These use cases don't get to + // aggregateMetrics because the new implementation is able to honor their table config setting + // of noDict on STRING/BYTES. Without metrics aggregation, memory pressure increases. + // So to continue aggregating metrics for such cases, we will create dictionary even + // if the column is part of noDictionary set from table config + if (fieldSpec instanceof DimensionFieldSpec && _aggregateMetrics && (dataType == FieldSpec.DataType.STRING || + dataType == FieldSpec.DataType.BYTES)) { + _logger.info("Not creating dictionary in consuming segment for column {} of type {}", column, dataType.toString()); + return false; + } + // So don't create dictionary if the column is member of noDictionary, is single-value + // and doesn't have an inverted index + return fieldSpec.isSingleValueField() && !invertedIndexColumns.contains(column); + } + // column is not a part of noDictionary set, so create dictionary + return false; } public SegmentPartitionConfig getSegmentPartitionConfig() { diff --git a/pinot-core/src/main/java/org/apache/pinot/core/segment/creator/impl/SegmentColumnarIndexCreator.java b/pinot-core/src/main/java/org/apache/pinot/core/segment/creator/impl/SegmentColumnarIndexCreator.java index 12e6feb..9ff2864 100644 --- a/pinot-core/src/main/java/org/apache/pinot/core/segment/creator/impl/SegmentColumnarIndexCreator.java +++ b/pinot-core/src/main/java/org/apache/pinot/core/segment/creator/impl/SegmentColumnarIndexCreator.java @@ -53,6 +53,7 @@ import org.apache.pinot.core.segment.creator.impl.inv.OffHeapBitmapInvertedIndex import org.apache.pinot.core.segment.creator.impl.inv.OnHeapBitmapInvertedIndexCreator; import org.apache.pinot.core.segment.creator.impl.inv.text.LuceneTextIndexCreator; import org.apache.pinot.core.segment.creator.impl.nullvalue.NullValueVectorCreator; +import org.apache.pinot.spi.config.table.FieldConfig; import org.apache.pinot.spi.data.DateTimeFieldSpec; import org.apache.pinot.spi.data.FieldSpec; import org.apache.pinot.spi.data.FieldSpec.FieldType; @@ -193,9 +194,10 @@ public class SegmentColumnarIndexCreator implements SegmentCreator { getColumnCompressionType(segmentCreationSpec, fieldSpec); // Initialize forward index creator + boolean deriveNumDocsPerChunk = shouldDeriveNumDocsPerChunk(columnName, segmentCreationSpec.getColumnProperties()); _forwardIndexCreatorMap.put(columnName, getRawIndexCreatorForColumn(_indexDir, compressionType, columnName, fieldSpec.getDataType(), totalDocs, - indexCreationInfo.getLengthOfLongestEntry())); + indexCreationInfo.getLengthOfLongestEntry(), deriveNumDocsPerChunk)); // Initialize text index creator if (_textIndexColumns.contains(columnName)) { @@ -213,6 +215,14 @@ public class SegmentColumnarIndexCreator implements SegmentCreator { } } + public static boolean shouldDeriveNumDocsPerChunk(String columnName, Map<String, Map<String, String>> columnProperties) { + if (columnProperties != null) { + Map<String, String> properties = columnProperties.get(columnName); + return properties != null && Boolean.parseBoolean(properties.get(FieldConfig.DERIVE_NUM_DOCS_PER_CHUNK_RAW_INDEX_KEY)); + } + return false; + } + /** * Helper method that returns compression type to use based on segment creation spec and field type. * <ul> @@ -539,6 +549,13 @@ public class SegmentColumnarIndexCreator implements SegmentCreator { ChunkCompressorFactory.CompressionType compressionType, String column, FieldSpec.DataType dataType, int totalDocs, int lengthOfLongestEntry) throws IOException { + return getRawIndexCreatorForColumn(file, compressionType, column, dataType, totalDocs, lengthOfLongestEntry, false); + } + + public static SingleValueRawIndexCreator getRawIndexCreatorForColumn(File file, + ChunkCompressorFactory.CompressionType compressionType, String column, FieldSpec.DataType dataType, int totalDocs, + int lengthOfLongestEntry, boolean deriveNumDocsPerChunk) + throws IOException { SingleValueRawIndexCreator indexCreator; switch (dataType) { @@ -561,7 +578,8 @@ public class SegmentColumnarIndexCreator implements SegmentCreator { case STRING: case BYTES: indexCreator = - new SingleValueVarByteRawIndexCreator(file, compressionType, column, totalDocs, lengthOfLongestEntry); + new SingleValueVarByteRawIndexCreator(file, compressionType, column, totalDocs, lengthOfLongestEntry, + deriveNumDocsPerChunk); break; default: diff --git a/pinot-core/src/main/java/org/apache/pinot/core/segment/creator/impl/fwd/SingleValueVarByteRawIndexCreator.java b/pinot-core/src/main/java/org/apache/pinot/core/segment/creator/impl/fwd/SingleValueVarByteRawIndexCreator.java index a8d7e6b..93652e5 100644 --- a/pinot-core/src/main/java/org/apache/pinot/core/segment/creator/impl/fwd/SingleValueVarByteRawIndexCreator.java +++ b/pinot-core/src/main/java/org/apache/pinot/core/segment/creator/impl/fwd/SingleValueVarByteRawIndexCreator.java @@ -28,6 +28,7 @@ import org.apache.pinot.core.segment.creator.impl.V1Constants; public class SingleValueVarByteRawIndexCreator extends BaseSingleValueRawIndexCreator { + private static final int DEFAULT_NUM_DOCS_PER_CHUNK = 1000; private static final int TARGET_MAX_CHUNK_SIZE = 1024 * 1024; private final VarByteChunkSingleValueWriter _indexWriter; @@ -35,8 +36,15 @@ public class SingleValueVarByteRawIndexCreator extends BaseSingleValueRawIndexCr public SingleValueVarByteRawIndexCreator(File baseIndexDir, ChunkCompressorFactory.CompressionType compressionType, String column, int totalDocs, int maxLength) throws IOException { + this(baseIndexDir, compressionType, column, totalDocs, maxLength, false); + } + + public SingleValueVarByteRawIndexCreator(File baseIndexDir, ChunkCompressorFactory.CompressionType compressionType, + String column, int totalDocs, int maxLength, boolean deriveNumDocsPerChunk) + throws IOException { File file = new File(baseIndexDir, column + V1Constants.Indexes.RAW_SV_FORWARD_INDEX_FILE_EXTENSION); - _indexWriter = new VarByteChunkSingleValueWriter(file, compressionType, totalDocs, getNumDocsPerChunk(maxLength), maxLength); + int numDocsPerChunk = deriveNumDocsPerChunk ? getNumDocsPerChunk(maxLength) : DEFAULT_NUM_DOCS_PER_CHUNK; + _indexWriter = new VarByteChunkSingleValueWriter(file, compressionType, totalDocs, numDocsPerChunk, maxLength); } @VisibleForTesting diff --git a/pinot-core/src/main/java/org/apache/pinot/core/segment/index/loader/defaultcolumn/BaseDefaultColumnHandler.java b/pinot-core/src/main/java/org/apache/pinot/core/segment/index/loader/defaultcolumn/BaseDefaultColumnHandler.java index d7277b9..12d1e9b 100644 --- a/pinot-core/src/main/java/org/apache/pinot/core/segment/index/loader/defaultcolumn/BaseDefaultColumnHandler.java +++ b/pinot-core/src/main/java/org/apache/pinot/core/segment/index/loader/defaultcolumn/BaseDefaultColumnHandler.java @@ -362,9 +362,10 @@ public abstract class BaseDefaultColumnHandler implements DefaultColumnHandler { int lengthOfLongestEntry = StringUtil.encodeUtf8(stringDefaultValue).length; int dictionaryElementSize = 0; + boolean deriveNumDocsPerChunk = SegmentColumnarIndexCreator.shouldDeriveNumDocsPerChunk(column, indexLoadingConfig.getColumnProperties()); SingleValueVarByteRawIndexCreator rawIndexCreator = new SingleValueVarByteRawIndexCreator(_indexDir, ChunkCompressorFactory.CompressionType.SNAPPY, column, - totalDocs, lengthOfLongestEntry); + totalDocs, lengthOfLongestEntry, deriveNumDocsPerChunk); for (int docId = 0; docId < totalDocs; docId++) { rawIndexCreator.index(docId, defaultValue); diff --git a/pinot-spi/src/main/java/org/apache/pinot/spi/config/table/FieldConfig.java b/pinot-spi/src/main/java/org/apache/pinot/spi/config/table/FieldConfig.java index 8e0d2e0..03d2590 100644 --- a/pinot-spi/src/main/java/org/apache/pinot/spi/config/table/FieldConfig.java +++ b/pinot-spi/src/main/java/org/apache/pinot/spi/config/table/FieldConfig.java @@ -36,6 +36,7 @@ public class FieldConfig extends BaseJsonConfig { public static String BLOOM_FILTER_COLUMN_KEY = "bloom.filter"; public static String ON_HEAP_DICTIONARY_COLUMN_KEY = "onheap.dictionary"; public static String VAR_LENGTH_DICTIONARY_COLUMN_KEY = "var.length.dictionary"; + public static String DERIVE_NUM_DOCS_PER_CHUNK_RAW_INDEX_KEY = "derive.num.docs.per.chunk.raw.index"; public static String TEXT_INDEX_REALTIME_READER_REFRESH_KEY = "text.index.realtime.reader.refresh"; // Lucene creates a query result cache if this option is enabled --------------------------------------------------------------------- To unsubscribe, e-mail: commits-unsubscr...@pinot.apache.org For additional commands, e-mail: commits-h...@pinot.apache.org