Search in sources :

Example 16 with QueryableIndexSegment

use of io.druid.segment.QueryableIndexSegment in project druid by druid-io.

the class GroupByTypeInterfaceBenchmark method querySingleQueryableIndexFloatTwice.

@Benchmark
@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.MICROSECONDS)
public void querySingleQueryableIndexFloatTwice(Blackhole blackhole) throws Exception {
    QueryRunner<Row> runner = QueryBenchmarkUtil.makeQueryRunner(factory, "qIndex", new QueryableIndexSegment("qIndex", queryableIndexes.get(0)));
    List<Row> results = GroupByTypeInterfaceBenchmark.runQuery(factory, runner, floatQuery);
    for (Row result : results) {
        blackhole.consume(result);
    }
    runner = QueryBenchmarkUtil.makeQueryRunner(factory, "qIndex", new QueryableIndexSegment("qIndex", queryableIndexes.get(0)));
    results = GroupByTypeInterfaceBenchmark.runQuery(factory, runner, floatQuery);
    for (Row result : results) {
        blackhole.consume(result);
    }
}
Also used : QueryableIndexSegment(io.druid.segment.QueryableIndexSegment) InputRow(io.druid.data.input.InputRow) Row(io.druid.data.input.Row) BenchmarkMode(org.openjdk.jmh.annotations.BenchmarkMode) Benchmark(org.openjdk.jmh.annotations.Benchmark) OutputTimeUnit(org.openjdk.jmh.annotations.OutputTimeUnit)

Example 17 with QueryableIndexSegment

use of io.druid.segment.QueryableIndexSegment in project druid by druid-io.

the class GroupByTypeInterfaceBenchmark method querySingleQueryableIndexNumericOnly.

@Benchmark
@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.MICROSECONDS)
public void querySingleQueryableIndexNumericOnly(Blackhole blackhole) throws Exception {
    QueryRunner<Row> runner = QueryBenchmarkUtil.makeQueryRunner(factory, "qIndex", new QueryableIndexSegment("qIndex", queryableIndexes.get(0)));
    List<Row> results = GroupByTypeInterfaceBenchmark.runQuery(factory, runner, longFloatQuery);
    for (Row result : results) {
        blackhole.consume(result);
    }
}
Also used : QueryableIndexSegment(io.druid.segment.QueryableIndexSegment) InputRow(io.druid.data.input.InputRow) Row(io.druid.data.input.Row) BenchmarkMode(org.openjdk.jmh.annotations.BenchmarkMode) Benchmark(org.openjdk.jmh.annotations.Benchmark) OutputTimeUnit(org.openjdk.jmh.annotations.OutputTimeUnit)

Example 18 with QueryableIndexSegment

use of io.druid.segment.QueryableIndexSegment in project druid by druid-io.

the class AppenderatorImpl method bootstrapSinksFromDisk.

/**
   * Populate "sinks" and "sinkTimeline" with committed segments, and announce them with the segmentAnnouncer.
   *
   * @return persisted commit metadata
   */
private Object bootstrapSinksFromDisk() {
    Preconditions.checkState(sinks.isEmpty(), "Already bootstrapped?!");
    final File baseDir = tuningConfig.getBasePersistDirectory();
    if (!baseDir.exists()) {
        return null;
    }
    final File[] files = baseDir.listFiles();
    if (files == null) {
        return null;
    }
    final File commitFile = computeCommitFile();
    final Committed committed;
    try {
        if (commitFile.exists()) {
            committed = objectMapper.readValue(commitFile, Committed.class);
        } else {
            committed = Committed.nil();
        }
    } catch (Exception e) {
        throw new ISE(e, "Failed to read commitFile: %s", commitFile);
    }
    log.info("Loading sinks from[%s]: %s", baseDir, committed.getHydrants().keySet());
    for (File sinkDir : files) {
        final File identifierFile = new File(sinkDir, IDENTIFIER_FILE_NAME);
        if (!identifierFile.isFile()) {
            // No identifier in this sinkDir; it must not actually be a sink directory. Skip it.
            continue;
        }
        try {
            final SegmentIdentifier identifier = objectMapper.readValue(new File(sinkDir, "identifier.json"), SegmentIdentifier.class);
            final int committedHydrants = committed.getCommittedHydrants(identifier.getIdentifierAsString());
            if (committedHydrants <= 0) {
                log.info("Removing uncommitted sink at [%s]", sinkDir);
                FileUtils.deleteDirectory(sinkDir);
                continue;
            }
            // To avoid reading and listing of "merged" dir and other special files
            final File[] sinkFiles = sinkDir.listFiles(new FilenameFilter() {

                @Override
                public boolean accept(File dir, String fileName) {
                    return !(Ints.tryParse(fileName) == null);
                }
            });
            Arrays.sort(sinkFiles, new Comparator<File>() {

                @Override
                public int compare(File o1, File o2) {
                    return Ints.compare(Integer.parseInt(o1.getName()), Integer.parseInt(o2.getName()));
                }
            });
            List<FireHydrant> hydrants = Lists.newArrayList();
            for (File hydrantDir : sinkFiles) {
                final int hydrantNumber = Integer.parseInt(hydrantDir.getName());
                if (hydrantNumber >= committedHydrants) {
                    log.info("Removing uncommitted segment at [%s]", hydrantDir);
                    FileUtils.deleteDirectory(hydrantDir);
                } else {
                    log.info("Loading previously persisted segment at [%s]", hydrantDir);
                    if (hydrantNumber != hydrants.size()) {
                        throw new ISE("Missing hydrant [%,d] in sinkDir [%s].", hydrants.size(), sinkDir);
                    }
                    hydrants.add(new FireHydrant(new QueryableIndexSegment(identifier.getIdentifierAsString(), indexIO.loadIndex(hydrantDir)), hydrantNumber));
                }
            }
            // Make sure we loaded enough hydrants.
            if (committedHydrants != hydrants.size()) {
                throw new ISE("Missing hydrant [%,d] in sinkDir [%s].", hydrants.size(), sinkDir);
            }
            Sink currSink = new Sink(identifier.getInterval(), schema, identifier.getShardSpec(), identifier.getVersion(), tuningConfig.getMaxRowsInMemory(), tuningConfig.isReportParseExceptions(), hydrants);
            sinks.put(identifier, currSink);
            sinkTimeline.add(currSink.getInterval(), currSink.getVersion(), identifier.getShardSpec().createChunk(currSink));
            segmentAnnouncer.announceSegment(currSink.getSegment());
        } catch (IOException e) {
            log.makeAlert(e, "Problem loading sink[%s] from disk.", schema.getDataSource()).addData("sinkDir", sinkDir).emit();
        }
    }
    // Make sure we loaded all committed sinks.
    final Set<String> loadedSinks = Sets.newHashSet(Iterables.transform(sinks.keySet(), new Function<SegmentIdentifier, String>() {

        @Override
        public String apply(SegmentIdentifier input) {
            return input.getIdentifierAsString();
        }
    }));
    final Set<String> missingSinks = Sets.difference(committed.getHydrants().keySet(), loadedSinks);
    if (!missingSinks.isEmpty()) {
        throw new ISE("Missing committed sinks [%s]", Joiner.on(", ").join(missingSinks));
    }
    return committed.getMetadata();
}
Also used : QueryableIndexSegment(io.druid.segment.QueryableIndexSegment) IOException(java.io.IOException) IndexSizeExceededException(io.druid.segment.incremental.IndexSizeExceededException) IOException(java.io.IOException) ExecutionException(java.util.concurrent.ExecutionException) FilenameFilter(java.io.FilenameFilter) Function(com.google.common.base.Function) Sink(io.druid.segment.realtime.plumber.Sink) ISE(io.druid.java.util.common.ISE) FireHydrant(io.druid.segment.realtime.FireHydrant) File(java.io.File)

Example 19 with QueryableIndexSegment

use of io.druid.segment.QueryableIndexSegment in project druid by druid-io.

the class SearchBenchmark method queryMultiQueryableIndex.

@Benchmark
@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.MICROSECONDS)
public void queryMultiQueryableIndex(Blackhole blackhole) throws Exception {
    List<QueryRunner<Row>> singleSegmentRunners = Lists.newArrayList();
    QueryToolChest toolChest = factory.getToolchest();
    for (int i = 0; i < numSegments; i++) {
        String segmentName = "qIndex" + i;
        final QueryRunner<Result<SearchResultValue>> runner = QueryBenchmarkUtil.makeQueryRunner(factory, segmentName, new QueryableIndexSegment(segmentName, qIndexes.get(i)));
        singleSegmentRunners.add(toolChest.preMergeQueryDecoration(runner));
    }
    QueryRunner theRunner = toolChest.postMergeQueryDecoration(new FinalizeResultsQueryRunner<>(toolChest.mergeResults(factory.mergeRunners(executorService, singleSegmentRunners)), toolChest));
    Sequence<Result<SearchResultValue>> queryResult = theRunner.run(query, Maps.<String, Object>newHashMap());
    List<Result<SearchResultValue>> results = Sequences.toList(queryResult, Lists.<Result<SearchResultValue>>newArrayList());
    for (Result<SearchResultValue> result : results) {
        List<SearchHit> hits = result.getValue().getValue();
        for (SearchHit hit : hits) {
            blackhole.consume(hit);
        }
    }
}
Also used : QueryableIndexSegment(io.druid.segment.QueryableIndexSegment) SearchHit(io.druid.query.search.search.SearchHit) SearchQueryQueryToolChest(io.druid.query.search.SearchQueryQueryToolChest) QueryToolChest(io.druid.query.QueryToolChest) FinalizeResultsQueryRunner(io.druid.query.FinalizeResultsQueryRunner) QueryRunner(io.druid.query.QueryRunner) Result(io.druid.query.Result) SearchResultValue(io.druid.query.search.SearchResultValue) BenchmarkMode(org.openjdk.jmh.annotations.BenchmarkMode) Benchmark(org.openjdk.jmh.annotations.Benchmark) OutputTimeUnit(org.openjdk.jmh.annotations.OutputTimeUnit)

Example 20 with QueryableIndexSegment

use of io.druid.segment.QueryableIndexSegment in project druid by druid-io.

the class SelectBenchmark method queryMultiQueryableIndex.

@Benchmark
@BenchmarkMode(Mode.AverageTime)
@OutputTimeUnit(TimeUnit.MICROSECONDS)
public void queryMultiQueryableIndex(Blackhole blackhole) throws Exception {
    SelectQuery queryCopy = query.withPagingSpec(PagingSpec.newSpec(pagingThreshold));
    String segmentName;
    List<QueryRunner<Result<SelectResultValue>>> singleSegmentRunners = Lists.newArrayList();
    QueryToolChest toolChest = factory.getToolchest();
    for (int i = 0; i < numSegments; i++) {
        segmentName = "qIndex" + i;
        QueryRunner<Result<SelectResultValue>> runner = QueryBenchmarkUtil.makeQueryRunner(factory, segmentName, new QueryableIndexSegment(segmentName, qIndexes.get(i)));
        singleSegmentRunners.add(toolChest.preMergeQueryDecoration(runner));
    }
    QueryRunner theRunner = toolChest.postMergeQueryDecoration(new FinalizeResultsQueryRunner<>(toolChest.mergeResults(factory.mergeRunners(executorService, singleSegmentRunners)), toolChest));
    boolean done = false;
    while (!done) {
        Sequence<Result<SelectResultValue>> queryResult = theRunner.run(queryCopy, Maps.<String, Object>newHashMap());
        List<Result<SelectResultValue>> results = Sequences.toList(queryResult, Lists.<Result<SelectResultValue>>newArrayList());
        SelectResultValue result = results.get(0).getValue();
        if (result.getEvents().size() == 0) {
            done = true;
        } else {
            for (EventHolder eh : result.getEvents()) {
                blackhole.consume(eh);
            }
            queryCopy = incrementQueryPagination(queryCopy, result);
        }
    }
}
Also used : QueryableIndexSegment(io.druid.segment.QueryableIndexSegment) SelectResultValue(io.druid.query.select.SelectResultValue) SelectQueryQueryToolChest(io.druid.query.select.SelectQueryQueryToolChest) QueryToolChest(io.druid.query.QueryToolChest) FinalizeResultsQueryRunner(io.druid.query.FinalizeResultsQueryRunner) QueryRunner(io.druid.query.QueryRunner) Result(io.druid.query.Result) SelectQuery(io.druid.query.select.SelectQuery) EventHolder(io.druid.query.select.EventHolder) BenchmarkMode(org.openjdk.jmh.annotations.BenchmarkMode) Benchmark(org.openjdk.jmh.annotations.Benchmark) OutputTimeUnit(org.openjdk.jmh.annotations.OutputTimeUnit)

Aggregations

QueryableIndexSegment (io.druid.segment.QueryableIndexSegment)56 Benchmark (org.openjdk.jmh.annotations.Benchmark)37 BenchmarkMode (org.openjdk.jmh.annotations.BenchmarkMode)37 OutputTimeUnit (org.openjdk.jmh.annotations.OutputTimeUnit)37 Result (io.druid.query.Result)22 Row (io.druid.data.input.Row)19 InputRow (io.druid.data.input.InputRow)16 TopNResultValue (io.druid.query.topn.TopNResultValue)15 IncrementalIndexSegment (io.druid.segment.IncrementalIndexSegment)9 QueryRunner (io.druid.query.QueryRunner)7 FinalizeResultsQueryRunner (io.druid.query.FinalizeResultsQueryRunner)5 DefaultDimensionSpec (io.druid.query.dimension.DefaultDimensionSpec)5 QueryableIndex (io.druid.segment.QueryableIndex)5 Test (org.junit.Test)5 QueryToolChest (io.druid.query.QueryToolChest)4 CountAggregatorFactory (io.druid.query.aggregation.CountAggregatorFactory)4 SelectorDimFilter (io.druid.query.filter.SelectorDimFilter)4 GroupByQueryRunnerTest (io.druid.query.groupby.GroupByQueryRunnerTest)4 TimeseriesResultValue (io.druid.query.timeseries.TimeseriesResultValue)4 IndexSpec (io.druid.segment.IndexSpec)4