Examples with ReferenceCountingSegment org.apache.druid.segment.ReferenceCountingSegment used on opensource projects

Example 6 with ReferenceCountingSegment

use of org.apache.druid.segment.ReferenceCountingSegment in project druid by druid-io.

the class StreamAppenderator method mergeAndPush.

/**
 * Merge segment, push to deep storage. Should only be used on segments that have been fully persisted. Must only
 * be run in the single-threaded pushExecutor.
 *
 * @param identifier    sink identifier
 * @param sink          sink to push
 * @param useUniquePath true if the segment should be written to a path with a unique identifier
 *
 * @return segment descriptor, or null if the sink is no longer valid
 */
@Nullable
private DataSegment mergeAndPush(final SegmentIdWithShardSpec identifier, final Sink sink, final boolean useUniquePath) {
    // noinspection ObjectEquality
    if (sinks.get(identifier) != sink) {
        log.warn("Sink for segment[%s] no longer valid, bailing out of mergeAndPush.", identifier);
        return null;
    }
    // Use a descriptor file to indicate that pushing has completed.
    final File persistDir = computePersistDir(identifier);
    final File mergedTarget = new File(persistDir, "merged");
    final File descriptorFile = computeDescriptorFile(identifier);
    // Sanity checks
    for (FireHydrant hydrant : sink) {
        if (sink.isWritable()) {
            throw new ISE("Expected sink to be no longer writable before mergeAndPush for segment[%s].", identifier);
        }
        synchronized (hydrant) {
            if (!hydrant.hasSwapped()) {
                throw new ISE("Expected sink to be fully persisted before mergeAndPush for segment[%s].", identifier);
            }
        }
    }
    try {
        if (descriptorFile.exists()) {
            if (useUniquePath) {
                // Don't reuse the descriptor, because the caller asked for a unique path. Leave the old one as-is, since
                // it might serve some unknown purpose.
                log.debug("Segment[%s] already pushed, but we want a unique path, so will push again with a new path.", identifier);
            } else {
                log.info("Segment[%s] already pushed, skipping.", identifier);
                return objectMapper.readValue(descriptorFile, DataSegment.class);
            }
        }
        removeDirectory(mergedTarget);
        if (mergedTarget.exists()) {
            throw new ISE("Merged target[%s] exists after removing?!", mergedTarget);
        }
        final File mergedFile;
        final long mergeFinishTime;
        final long startTime = System.nanoTime();
        List<QueryableIndex> indexes = new ArrayList<>();
        Closer closer = Closer.create();
        try {
            for (FireHydrant fireHydrant : sink) {
                Pair<ReferenceCountingSegment, Closeable> segmentAndCloseable = fireHydrant.getAndIncrementSegment();
                final QueryableIndex queryableIndex = segmentAndCloseable.lhs.asQueryableIndex();
                log.debug("Segment[%s] adding hydrant[%s]", identifier, fireHydrant);
                indexes.add(queryableIndex);
                closer.register(segmentAndCloseable.rhs);
            }
            mergedFile = indexMerger.mergeQueryableIndex(indexes, schema.getGranularitySpec().isRollup(), schema.getAggregators(), schema.getDimensionsSpec(), mergedTarget, tuningConfig.getIndexSpec(), tuningConfig.getIndexSpecForIntermediatePersists(), new BaseProgressIndicator(), tuningConfig.getSegmentWriteOutMediumFactory(), tuningConfig.getMaxColumnsToMerge());
            mergeFinishTime = System.nanoTime();
            log.debug("Segment[%s] built in %,dms.", identifier, (mergeFinishTime - startTime) / 1000000);
        } catch (Throwable t) {
            throw closer.rethrow(t);
        } finally {
            closer.close();
        }
        final DataSegment segmentToPush = sink.getSegment().withDimensions(IndexMerger.getMergedDimensionsFromQueryableIndexes(indexes, schema.getDimensionsSpec()));
        // Retry pushing segments because uploading to deep storage might fail especially for cloud storage types
        final DataSegment segment = RetryUtils.retry(// semantics.
        () -> dataSegmentPusher.push(mergedFile, segmentToPush, useUniquePath), exception -> exception instanceof Exception, 5);
        final long pushFinishTime = System.nanoTime();
        objectMapper.writeValue(descriptorFile, segment);
        log.info("Segment[%s] of %,d bytes " + "built from %d incremental persist(s) in %,dms; " + "pushed to deep storage in %,dms. " + "Load spec is: %s", identifier, segment.getSize(), indexes.size(), (mergeFinishTime - startTime) / 1000000, (pushFinishTime - mergeFinishTime) / 1000000, objectMapper.writeValueAsString(segment.getLoadSpec()));
        return segment;
    } catch (Exception e) {
        metrics.incrementFailedHandoffs();
        log.warn(e, "Failed to push merged index for segment[%s].", identifier);
        throw new RuntimeException(e);
    }
}

Example 7 with ReferenceCountingSegment

use of org.apache.druid.segment.ReferenceCountingSegment in project druid by druid-io.

the class Sink method makeNewCurrIndex.

private FireHydrant makeNewCurrIndex(long minTimestamp, DataSchema schema) {
    final IncrementalIndexSchema indexSchema = new IncrementalIndexSchema.Builder().withMinTimestamp(minTimestamp).withTimestampSpec(schema.getTimestampSpec()).withQueryGranularity(schema.getGranularitySpec().getQueryGranularity()).withDimensionsSpec(schema.getDimensionsSpec()).withMetrics(schema.getAggregators()).withRollup(schema.getGranularitySpec().isRollup()).build();
    // Build the incremental-index according to the spec that was chosen by the user
    final IncrementalIndex newIndex = appendableIndexSpec.builder().setIndexSchema(indexSchema).setMaxRowCount(maxRowsInMemory).setMaxBytesInMemory(maxBytesInMemory).setUseMaxMemoryEstimates(useMaxMemoryEstimates).build();
    final FireHydrant old;
    synchronized (hydrantLock) {
        if (writable) {
            old = currHydrant;
            int newCount = 0;
            int numHydrants = hydrants.size();
            if (numHydrants > 0) {
                FireHydrant lastHydrant = hydrants.get(numHydrants - 1);
                newCount = lastHydrant.getCount() + 1;
                if (!indexSchema.getDimensionsSpec().hasCustomDimensions()) {
                    Map<String, ColumnCapabilities> oldCapabilities;
                    if (lastHydrant.hasSwapped()) {
                        oldCapabilities = new HashMap<>();
                        ReferenceCountingSegment segment = lastHydrant.getIncrementedSegment();
                        try {
                            QueryableIndex oldIndex = segment.asQueryableIndex();
                            for (String dim : oldIndex.getAvailableDimensions()) {
                                dimOrder.add(dim);
                                oldCapabilities.put(dim, oldIndex.getColumnHolder(dim).getCapabilities());
                            }
                        } finally {
                            segment.decrement();
                        }
                    } else {
                        IncrementalIndex oldIndex = lastHydrant.getIndex();
                        dimOrder.addAll(oldIndex.getDimensionOrder());
                        oldCapabilities = oldIndex.getColumnCapabilities();
                    }
                    newIndex.loadDimensionIterable(dimOrder, oldCapabilities);
                }
            }
            currHydrant = new FireHydrant(newIndex, newCount, getSegment().getId());
            if (old != null) {
                numRowsExcludingCurrIndex.addAndGet(old.getIndex().size());
            }
            hydrants.add(currHydrant);
        } else {
            // Oops, someone called finishWriting while we were making this new index.
            newIndex.close();
            throw new ISE("finishWriting() called during swap");
        }
    }
    return old;
}

Example 8 with ReferenceCountingSegment

use of org.apache.druid.segment.ReferenceCountingSegment in project druid by druid-io.

the class RealtimePlumberSchoolTest method testDimOrderInheritanceHelper.

private void testDimOrderInheritanceHelper(final Object commitMetadata) throws Exception {
    List<List<String>> expectedDims = ImmutableList.of(ImmutableList.of("dimD"), ImmutableList.of("dimC"), ImmutableList.of("dimA"), ImmutableList.of("dimB"), ImmutableList.of("dimE"), ImmutableList.of("dimD", "dimC", "dimA", "dimB", "dimE"));
    QueryableIndex qindex;
    FireHydrant hydrant;
    Map<Long, Sink> sinks;
    RealtimePlumber plumber = (RealtimePlumber) realtimePlumberSchool.findPlumber(schema2, tuningConfig, metrics);
    Assert.assertNull(plumber.startJob());
    final CountDownLatch doneSignal = new CountDownLatch(1);
    final Committer committer = new Committer() {

        @Override
        public Object getMetadata() {
            return commitMetadata;
        }

        @Override
        public void run() {
            doneSignal.countDown();
        }
    };
    plumber.add(getTestInputRowFull("1970-01-01", ImmutableList.of("dimD"), ImmutableList.of("1")), Suppliers.ofInstance(committer));
    plumber.add(getTestInputRowFull("1970-01-01", ImmutableList.of("dimC"), ImmutableList.of("1")), Suppliers.ofInstance(committer));
    plumber.add(getTestInputRowFull("1970-01-01", ImmutableList.of("dimA"), ImmutableList.of("1")), Suppliers.ofInstance(committer));
    plumber.add(getTestInputRowFull("1970-01-01", ImmutableList.of("dimB"), ImmutableList.of("1")), Suppliers.ofInstance(committer));
    plumber.add(getTestInputRowFull("1970-01-01", ImmutableList.of("dimE"), ImmutableList.of("1")), Suppliers.ofInstance(committer));
    plumber.add(getTestInputRowFull("1970-01-01", ImmutableList.of("dimA", "dimB", "dimC", "dimD", "dimE"), ImmutableList.of("1")), Suppliers.ofInstance(committer));
    plumber.persist(committer);
    doneSignal.await();
    plumber.getSinks().clear();
    plumber.finishJob();
    RealtimePlumber restoredPlumber = (RealtimePlumber) realtimePlumberSchool.findPlumber(schema2, tuningConfig, metrics);
    restoredPlumber.bootstrapSinksFromDisk();
    sinks = restoredPlumber.getSinks();
    Assert.assertEquals(1, sinks.size());
    List<FireHydrant> hydrants = Lists.newArrayList(sinks.get(0L));
    for (int i = 0; i < hydrants.size(); i++) {
        hydrant = hydrants.get(i);
        ReferenceCountingSegment segment = hydrant.getIncrementedSegment();
        try {
            qindex = segment.asQueryableIndex();
            Assert.assertEquals(i, hydrant.getCount());
            Assert.assertEquals(expectedDims.get(i), ImmutableList.copyOf(qindex.getAvailableDimensions()));
        } finally {
            segment.decrement();
        }
    }
}

Example 9 with ReferenceCountingSegment

use of org.apache.druid.segment.ReferenceCountingSegment in project druid by druid-io.

the class ClientQuerySegmentWalkerTest method initWalker.

/**
 * Initialize (or reinitialize) our {@link #walker} and {@link #closer}.
 */
private void initWalker(final Map<String, String> serverProperties, QueryScheduler schedulerForTest) {
    final ObjectMapper jsonMapper = TestHelper.makeJsonMapper();
    final ServerConfig serverConfig = jsonMapper.convertValue(serverProperties, ServerConfig.class);
    final SegmentWrangler segmentWrangler = new MapSegmentWrangler(ImmutableMap.<Class<? extends DataSource>, SegmentWrangler>builder().put(InlineDataSource.class, new InlineSegmentWrangler()).build());
    final JoinableFactory globalFactory = new JoinableFactory() {

        @Override
        public boolean isDirectlyJoinable(DataSource dataSource) {
            return ((GlobalTableDataSource) dataSource).getName().equals(GLOBAL);
        }

        @Override
        public Optional<Joinable> build(DataSource dataSource, JoinConditionAnalysis condition) {
            return Optional.empty();
        }
    };
    final JoinableFactory joinableFactory = new MapJoinableFactory(ImmutableSet.of(globalFactory, new InlineJoinableFactory()), ImmutableMap.<Class<? extends JoinableFactory>, Class<? extends DataSource>>builder().put(InlineJoinableFactory.class, InlineDataSource.class).put(globalFactory.getClass(), GlobalTableDataSource.class).build());
    class CapturingWalker implements QuerySegmentWalker {

        private QuerySegmentWalker baseWalker;

        private ClusterOrLocal how;

        CapturingWalker(QuerySegmentWalker baseWalker, ClusterOrLocal how) {
            this.baseWalker = baseWalker;
            this.how = how;
        }

        @Override
        public <T> QueryRunner<T> getQueryRunnerForIntervals(Query<T> query, Iterable<Interval> intervals) {
            final QueryRunner<T> baseRunner = baseWalker.getQueryRunnerForIntervals(query, intervals);
            return (queryPlus, responseContext) -> {
                log.info("Query (%s): %s", how, queryPlus.getQuery());
                issuedQueries.add(new ExpectedQuery(queryPlus.getQuery(), how));
                return baseRunner.run(queryPlus, responseContext);
            };
        }

        @Override
        public <T> QueryRunner<T> getQueryRunnerForSegments(Query<T> query, Iterable<SegmentDescriptor> specs) {
            final QueryRunner<T> baseRunner = baseWalker.getQueryRunnerForSegments(query, specs);
            return (queryPlus, responseContext) -> {
                log.info("Query (%s): %s", how, queryPlus.getQuery());
                issuedQueries.add(new ExpectedQuery(queryPlus.getQuery(), how));
                return baseRunner.run(queryPlus, responseContext);
            };
        }
    }
    walker = QueryStackTests.createClientQuerySegmentWalker(new CapturingWalker(QueryStackTests.createClusterQuerySegmentWalker(ImmutableMap.<String, VersionedIntervalTimeline<String, ReferenceCountingSegment>>builder().put(FOO, makeTimeline(FOO, FOO_INLINE)).put(BAR, makeTimeline(BAR, BAR_INLINE)).put(MULTI, makeTimeline(MULTI, MULTI_VALUE_INLINE)).put(GLOBAL, makeTimeline(GLOBAL, FOO_INLINE)).put(ARRAY, makeTimeline(ARRAY, ARRAY_INLINE)).put(ARRAY_UNKNOWN, makeTimeline(ARRAY_UNKNOWN, ARRAY_INLINE_UNKNOWN)).build(), joinableFactory, conglomerate, schedulerForTest), ClusterOrLocal.CLUSTER), new CapturingWalker(QueryStackTests.createLocalQuerySegmentWalker(conglomerate, segmentWrangler, joinableFactory, schedulerForTest), ClusterOrLocal.LOCAL), conglomerate, joinableFactory, serverConfig);
}

Example 10 with ReferenceCountingSegment

use of org.apache.druid.segment.ReferenceCountingSegment in project druid by druid-io.

the class SpecificSegmentsQuerySegmentWalker method add.

public SpecificSegmentsQuerySegmentWalker add(final DataSegment descriptor, final QueryableIndex index) {
    final ReferenceCountingSegment segment = ReferenceCountingSegment.wrapSegment(new QueryableIndexSegment(index, descriptor.getId()), descriptor.getShardSpec());
    final VersionedIntervalTimeline<String, ReferenceCountingSegment> timeline = timelines.computeIfAbsent(descriptor.getDataSource(), datasource -> new VersionedIntervalTimeline<>(Ordering.natural()));
    timeline.add(descriptor.getInterval(), descriptor.getVersion(), descriptor.getShardSpec().createChunk(segment));
    segments.add(descriptor);
    closeables.add(segment);
    return this;
}