Examples with ForegroundCachePopulator org.apache.druid.client.cache.ForegroundCachePopulator used on opensource projects

Example 6 with ForegroundCachePopulator

use of org.apache.druid.client.cache.ForegroundCachePopulator in project druid by druid-io.

the class CachingClusteredClientBenchmark method setup.

@Setup(Level.Trial)
public void setup() {
    final String schemaName = "basic";
    parallelCombine = parallelism > 0;
    GeneratorSchemaInfo schemaInfo = GeneratorBasicSchemas.SCHEMA_MAP.get(schemaName);
    Map<DataSegment, QueryableIndex> queryableIndexes = Maps.newHashMapWithExpectedSize(numServers);
    for (int i = 0; i < numServers; i++) {
        final DataSegment dataSegment = DataSegment.builder().dataSource(DATA_SOURCE).interval(schemaInfo.getDataInterval()).version("1").shardSpec(new LinearShardSpec(i)).size(0).build();
        final SegmentGenerator segmentGenerator = closer.register(new SegmentGenerator());
        LOG.info("Starting benchmark setup using cacheDir[%s], rows[%,d].", segmentGenerator.getCacheDir(), rowsPerSegment);
        final QueryableIndex index = segmentGenerator.generate(dataSegment, schemaInfo, Granularities.NONE, rowsPerSegment);
        queryableIndexes.put(dataSegment, index);
    }
    final DruidProcessingConfig processingConfig = new DruidProcessingConfig() {

        @Override
        public String getFormatString() {
            return null;
        }

        @Override
        public int intermediateComputeSizeBytes() {
            return PROCESSING_BUFFER_SIZE;
        }

        @Override
        public int getNumMergeBuffers() {
            return 1;
        }

        @Override
        public int getNumThreads() {
            return numProcessingThreads;
        }

        @Override
        public boolean useParallelMergePool() {
            return true;
        }
    };
    conglomerate = new DefaultQueryRunnerFactoryConglomerate(ImmutableMap.<Class<? extends Query>, QueryRunnerFactory>builder().put(TimeseriesQuery.class, new TimeseriesQueryRunnerFactory(new TimeseriesQueryQueryToolChest(), new TimeseriesQueryEngine(), QueryRunnerTestHelper.NOOP_QUERYWATCHER)).put(TopNQuery.class, new TopNQueryRunnerFactory(new StupidPool<>("TopNQueryRunnerFactory-bufferPool", () -> ByteBuffer.allocate(PROCESSING_BUFFER_SIZE)), new TopNQueryQueryToolChest(new TopNQueryConfig()), QueryRunnerTestHelper.NOOP_QUERYWATCHER)).put(GroupByQuery.class, makeGroupByQueryRunnerFactory(GroupByQueryRunnerTest.DEFAULT_MAPPER, new GroupByQueryConfig() {

        @Override
        public String getDefaultStrategy() {
            return GroupByStrategySelector.STRATEGY_V2;
        }
    }, processingConfig)).build());
    toolChestWarehouse = new QueryToolChestWarehouse() {

        @Override
        public <T, QueryType extends Query<T>> QueryToolChest<T, QueryType> getToolChest(final QueryType query) {
            return conglomerate.findFactory(query).getToolchest();
        }
    };
    SimpleServerView serverView = new SimpleServerView();
    int serverSuffx = 1;
    for (Entry<DataSegment, QueryableIndex> entry : queryableIndexes.entrySet()) {
        serverView.addServer(createServer(serverSuffx++), entry.getKey(), entry.getValue());
    }
    processingPool = Execs.multiThreaded(processingConfig.getNumThreads(), "caching-clustered-client-benchmark");
    forkJoinPool = new ForkJoinPool((int) Math.ceil(Runtime.getRuntime().availableProcessors() * 0.75), ForkJoinPool.defaultForkJoinWorkerThreadFactory, null, true);
    cachingClusteredClient = new CachingClusteredClient(toolChestWarehouse, serverView, MapCache.create(0), JSON_MAPPER, new ForegroundCachePopulator(JSON_MAPPER, new CachePopulatorStats(), 0), new CacheConfig(), new DruidHttpClientConfig(), processingConfig, forkJoinPool, QueryStackTests.DEFAULT_NOOP_SCHEDULER, new MapJoinableFactory(ImmutableSet.of(), ImmutableMap.of()), new NoopServiceEmitter());
}

Example 7 with ForegroundCachePopulator

use of org.apache.druid.client.cache.ForegroundCachePopulator in project druid by druid-io.

the class SinkQuerySegmentWalker method getQueryRunnerForSegments.

@Override
public <T> QueryRunner<T> getQueryRunnerForSegments(final Query<T> query, final Iterable<SegmentDescriptor> specs) {
    // We only handle one particular dataSource. Make sure that's what we have, then ignore from here on out.
    final DataSourceAnalysis analysis = DataSourceAnalysis.forDataSource(query.getDataSource());
    // Sanity check: make sure the query is based on the table we're meant to handle.
    if (!analysis.getBaseTableDataSource().filter(ds -> dataSource.equals(ds.getName())).isPresent()) {
        throw new ISE("Cannot handle datasource: %s", analysis.getDataSource());
    }
    final QueryRunnerFactory<T, Query<T>> factory = conglomerate.findFactory(query);
    if (factory == null) {
        throw new ISE("Unknown query type[%s].", query.getClass());
    }
    final QueryToolChest<T, Query<T>> toolChest = factory.getToolchest();
    final boolean skipIncrementalSegment = query.getContextValue(CONTEXT_SKIP_INCREMENTAL_SEGMENT, false);
    final AtomicLong cpuTimeAccumulator = new AtomicLong(0L);
    // Make sure this query type can handle the subquery, if present.
    if (analysis.isQuery() && !toolChest.canPerformSubquery(((QueryDataSource) analysis.getDataSource()).getQuery())) {
        throw new ISE("Cannot handle subquery: %s", analysis.getDataSource());
    }
    // segmentMapFn maps each base Segment into a joined Segment if necessary.
    final Function<SegmentReference, SegmentReference> segmentMapFn = joinableFactoryWrapper.createSegmentMapFn(analysis.getJoinBaseTableFilter().map(Filters::toFilter).orElse(null), analysis.getPreJoinableClauses(), cpuTimeAccumulator, analysis.getBaseQuery().orElse(query));
    // We compute the join cache key here itself so it doesn't need to be re-computed for every segment
    final Optional<byte[]> cacheKeyPrefix = analysis.isJoin() ? joinableFactoryWrapper.computeJoinDataSourceCacheKey(analysis) : Optional.of(StringUtils.EMPTY_BYTES);
    Iterable<QueryRunner<T>> perSegmentRunners = Iterables.transform(specs, descriptor -> {
        final PartitionChunk<Sink> chunk = sinkTimeline.findChunk(descriptor.getInterval(), descriptor.getVersion(), descriptor.getPartitionNumber());
        if (chunk == null) {
            return new ReportTimelineMissingSegmentQueryRunner<>(descriptor);
        }
        final Sink theSink = chunk.getObject();
        final SegmentId sinkSegmentId = theSink.getSegment().getId();
        Iterable<QueryRunner<T>> perHydrantRunners = new SinkQueryRunners<>(Iterables.transform(theSink, hydrant -> {
            // Hydrant might swap at any point, but if it's swapped at the start
            // then we know it's *definitely* swapped.
            final boolean hydrantDefinitelySwapped = hydrant.hasSwapped();
            if (skipIncrementalSegment && !hydrantDefinitelySwapped) {
                return new Pair<>(hydrant.getSegmentDataInterval(), new NoopQueryRunner<>());
            }
            // Prevent the underlying segment from swapping when its being iterated
            final Optional<Pair<SegmentReference, Closeable>> maybeSegmentAndCloseable = hydrant.getSegmentForQuery(segmentMapFn);
            // if optional isn't present, we failed to acquire reference to the segment or any joinables
            if (!maybeSegmentAndCloseable.isPresent()) {
                return new Pair<>(hydrant.getSegmentDataInterval(), new ReportTimelineMissingSegmentQueryRunner<>(descriptor));
            }
            final Pair<SegmentReference, Closeable> segmentAndCloseable = maybeSegmentAndCloseable.get();
            try {
                QueryRunner<T> runner = factory.createRunner(segmentAndCloseable.lhs);
                // 2) Hydrants are not the same between replicas, make sure cache is local
                if (hydrantDefinitelySwapped && cache.isLocal()) {
                    StorageAdapter storageAdapter = segmentAndCloseable.lhs.asStorageAdapter();
                    long segmentMinTime = storageAdapter.getMinTime().getMillis();
                    long segmentMaxTime = storageAdapter.getMaxTime().getMillis();
                    Interval actualDataInterval = Intervals.utc(segmentMinTime, segmentMaxTime + 1);
                    runner = new CachingQueryRunner<>(makeHydrantCacheIdentifier(hydrant), cacheKeyPrefix, descriptor, actualDataInterval, objectMapper, cache, toolChest, runner, // Always populate in foreground regardless of config
                    new ForegroundCachePopulator(objectMapper, cachePopulatorStats, cacheConfig.getMaxEntrySize()), cacheConfig);
                }
                // Make it always use Closeable to decrement()
                runner = QueryRunnerHelper.makeClosingQueryRunner(runner, segmentAndCloseable.rhs);
                return new Pair<>(segmentAndCloseable.lhs.getDataInterval(), runner);
            } catch (Throwable e) {
                throw CloseableUtils.closeAndWrapInCatch(e, segmentAndCloseable.rhs);
            }
        }));
        return new SpecificSegmentQueryRunner<>(withPerSinkMetrics(new BySegmentQueryRunner<>(sinkSegmentId, descriptor.getInterval().getStart(), factory.mergeRunners(DirectQueryProcessingPool.INSTANCE, perHydrantRunners)), toolChest, sinkSegmentId, cpuTimeAccumulator), new SpecificSegmentSpec(descriptor));
    });
    final QueryRunner<T> mergedRunner = toolChest.mergeResults(factory.mergeRunners(queryProcessingPool, perSegmentRunners));
    return CPUTimeMetricQueryRunner.safeBuild(new FinalizeResultsQueryRunner<>(mergedRunner, toolChest), toolChest, emitter, cpuTimeAccumulator, true);
}

Example 8 with ForegroundCachePopulator

use of org.apache.druid.client.cache.ForegroundCachePopulator in project druid by druid-io.

the class CachingClusteredClientFunctionalityTest method setUp.

@Before
public void setUp() {
    timeline = new VersionedIntervalTimeline<>(Ordering.natural());
    serverView = EasyMock.createNiceMock(TimelineServerView.class);
    cache = MapCache.create(100000);
    client = makeClient(new ForegroundCachePopulator(OBJECT_MAPPER, new CachePopulatorStats(), -1));
}