use of com.google.common.base.Supplier in project druid by druid-io.
the class AppenderatorTest method testRestoreFromDisk.
@Test
public void testRestoreFromDisk() throws Exception {
final RealtimeTuningConfig tuningConfig;
try (final AppenderatorTester tester = new AppenderatorTester(2)) {
final Appenderator appenderator = tester.getAppenderator();
tuningConfig = tester.getTuningConfig();
final AtomicInteger eventCount = new AtomicInteger(0);
final Supplier<Committer> committerSupplier = new Supplier<Committer>() {
@Override
public Committer get() {
final Object metadata = ImmutableMap.of("eventCount", eventCount.get());
return new Committer() {
@Override
public Object getMetadata() {
return metadata;
}
@Override
public void run() {
// Do nothing
}
};
}
};
appenderator.startJob();
eventCount.incrementAndGet();
appenderator.add(IDENTIFIERS.get(0), IR("2000", "foo", 1), committerSupplier);
eventCount.incrementAndGet();
appenderator.add(IDENTIFIERS.get(0), IR("2000", "bar", 2), committerSupplier);
eventCount.incrementAndGet();
appenderator.add(IDENTIFIERS.get(0), IR("2000", "baz", 3), committerSupplier);
eventCount.incrementAndGet();
appenderator.add(IDENTIFIERS.get(0), IR("2000", "qux", 4), committerSupplier);
eventCount.incrementAndGet();
appenderator.add(IDENTIFIERS.get(0), IR("2000", "bob", 5), committerSupplier);
appenderator.close();
try (final AppenderatorTester tester2 = new AppenderatorTester(2, tuningConfig.getBasePersistDirectory())) {
final Appenderator appenderator2 = tester2.getAppenderator();
Assert.assertEquals(ImmutableMap.of("eventCount", 4), appenderator2.startJob());
Assert.assertEquals(ImmutableList.of(IDENTIFIERS.get(0)), appenderator2.getSegments());
Assert.assertEquals(4, appenderator2.getRowCount(IDENTIFIERS.get(0)));
}
}
}
use of com.google.common.base.Supplier in project druid by druid-io.
the class AppenderatorTest method testMaxRowsInMemory.
@Test
public void testMaxRowsInMemory() throws Exception {
try (final AppenderatorTester tester = new AppenderatorTester(3)) {
final Appenderator appenderator = tester.getAppenderator();
final AtomicInteger eventCount = new AtomicInteger(0);
final Supplier<Committer> committerSupplier = new Supplier<Committer>() {
@Override
public Committer get() {
final Object metadata = ImmutableMap.of("eventCount", eventCount.get());
return new Committer() {
@Override
public Object getMetadata() {
return metadata;
}
@Override
public void run() {
// Do nothing
}
};
}
};
Assert.assertEquals(0, ((AppenderatorImpl) appenderator).getRowsInMemory());
appenderator.startJob();
Assert.assertEquals(0, ((AppenderatorImpl) appenderator).getRowsInMemory());
appenderator.add(IDENTIFIERS.get(0), IR("2000", "foo", 1), committerSupplier);
Assert.assertEquals(1, ((AppenderatorImpl) appenderator).getRowsInMemory());
appenderator.add(IDENTIFIERS.get(1), IR("2000", "bar", 1), committerSupplier);
Assert.assertEquals(2, ((AppenderatorImpl) appenderator).getRowsInMemory());
appenderator.add(IDENTIFIERS.get(1), IR("2000", "bar", 1), committerSupplier);
Assert.assertEquals(2, ((AppenderatorImpl) appenderator).getRowsInMemory());
appenderator.add(IDENTIFIERS.get(0), IR("2000", "baz", 1), committerSupplier);
Assert.assertEquals(0, ((AppenderatorImpl) appenderator).getRowsInMemory());
appenderator.add(IDENTIFIERS.get(1), IR("2000", "qux", 1), committerSupplier);
Assert.assertEquals(1, ((AppenderatorImpl) appenderator).getRowsInMemory());
appenderator.add(IDENTIFIERS.get(0), IR("2000", "bob", 1), committerSupplier);
Assert.assertEquals(2, ((AppenderatorImpl) appenderator).getRowsInMemory());
appenderator.close();
Assert.assertEquals(0, ((AppenderatorImpl) appenderator).getRowsInMemory());
}
}
use of com.google.common.base.Supplier in project druid by druid-io.
the class DruidPlanner method planWithBindableConvention.
private PlannerResult planWithBindableConvention(final SqlExplain explain, final RelRoot root) throws RelConversionException {
BindableRel bindableRel = (BindableRel) planner.transform(Rules.BINDABLE_CONVENTION_RULES, planner.getEmptyTraitSet().replace(BindableConvention.INSTANCE).plus(root.collation), root.rel);
if (!root.isRefTrivial()) {
// Add a projection on top to accommodate root.fields.
final List<RexNode> projects = new ArrayList<>();
final RexBuilder rexBuilder = bindableRel.getCluster().getRexBuilder();
for (int field : Pair.left(root.fields)) {
projects.add(rexBuilder.makeInputRef(bindableRel, field));
}
bindableRel = new Bindables.BindableProject(bindableRel.getCluster(), bindableRel.getTraitSet(), bindableRel, projects, root.validatedRowType);
}
if (explain != null) {
return planExplanation(bindableRel, explain);
} else {
final BindableRel theRel = bindableRel;
final DataContext dataContext = plannerContext.createDataContext((JavaTypeFactory) planner.getTypeFactory());
final Supplier<Sequence<Object[]>> resultsSupplier = new Supplier<Sequence<Object[]>>() {
@Override
public Sequence<Object[]> get() {
final Enumerable enumerable = theRel.bind(dataContext);
return Sequences.simple(enumerable);
}
};
return new PlannerResult(resultsSupplier, root.validatedRowType);
}
}
use of com.google.common.base.Supplier in project druid by druid-io.
the class KafkaIndexTask method run.
@Override
public TaskStatus run(final TaskToolbox toolbox) throws Exception {
log.info("Starting up!");
startTime = DateTime.now();
mapper = toolbox.getObjectMapper();
status = Status.STARTING;
if (chatHandlerProvider.isPresent()) {
log.info("Found chat handler of class[%s]", chatHandlerProvider.get().getClass().getName());
chatHandlerProvider.get().register(getId(), this, false);
} else {
log.warn("No chat handler detected");
}
runThread = Thread.currentThread();
// Set up FireDepartmentMetrics
final FireDepartment fireDepartmentForMetrics = new FireDepartment(dataSchema, new RealtimeIOConfig(null, null, null), null);
fireDepartmentMetrics = fireDepartmentForMetrics.getMetrics();
toolbox.getMonitorScheduler().addMonitor(new RealtimeMetricsMonitor(ImmutableList.of(fireDepartmentForMetrics), ImmutableMap.of(DruidMetrics.TASK_ID, new String[] { getId() })));
try (final Appenderator appenderator0 = newAppenderator(fireDepartmentMetrics, toolbox);
final FiniteAppenderatorDriver driver = newDriver(appenderator0, toolbox, fireDepartmentMetrics);
final KafkaConsumer<byte[], byte[]> consumer = newConsumer()) {
appenderator = appenderator0;
final String topic = ioConfig.getStartPartitions().getTopic();
// Start up, set up initial offsets.
final Object restoredMetadata = driver.startJob();
if (restoredMetadata == null) {
nextOffsets.putAll(ioConfig.getStartPartitions().getPartitionOffsetMap());
} else {
final Map<String, Object> restoredMetadataMap = (Map) restoredMetadata;
final KafkaPartitions restoredNextPartitions = toolbox.getObjectMapper().convertValue(restoredMetadataMap.get(METADATA_NEXT_PARTITIONS), KafkaPartitions.class);
nextOffsets.putAll(restoredNextPartitions.getPartitionOffsetMap());
// Sanity checks.
if (!restoredNextPartitions.getTopic().equals(ioConfig.getStartPartitions().getTopic())) {
throw new ISE("WTF?! Restored topic[%s] but expected topic[%s]", restoredNextPartitions.getTopic(), ioConfig.getStartPartitions().getTopic());
}
if (!nextOffsets.keySet().equals(ioConfig.getStartPartitions().getPartitionOffsetMap().keySet())) {
throw new ISE("WTF?! Restored partitions[%s] but expected partitions[%s]", nextOffsets.keySet(), ioConfig.getStartPartitions().getPartitionOffsetMap().keySet());
}
}
// Set up sequenceNames.
final Map<Integer, String> sequenceNames = Maps.newHashMap();
for (Integer partitionNum : nextOffsets.keySet()) {
sequenceNames.put(partitionNum, String.format("%s_%s", ioConfig.getBaseSequenceName(), partitionNum));
}
// Set up committer.
final Supplier<Committer> committerSupplier = new Supplier<Committer>() {
@Override
public Committer get() {
final Map<Integer, Long> snapshot = ImmutableMap.copyOf(nextOffsets);
return new Committer() {
@Override
public Object getMetadata() {
return ImmutableMap.of(METADATA_NEXT_PARTITIONS, new KafkaPartitions(ioConfig.getStartPartitions().getTopic(), snapshot));
}
@Override
public void run() {
// Do nothing.
}
};
}
};
Set<Integer> assignment = assignPartitionsAndSeekToNext(consumer, topic);
// Main loop.
// Could eventually support leader/follower mode (for keeping replicas more in sync)
boolean stillReading = !assignment.isEmpty();
status = Status.READING;
try {
while (stillReading) {
if (possiblyPause(assignment)) {
// The partition assignments may have changed while paused by a call to setEndOffsets() so reassign
// partitions upon resuming. This is safe even if the end offsets have not been modified.
assignment = assignPartitionsAndSeekToNext(consumer, topic);
if (assignment.isEmpty()) {
log.info("All partitions have been fully read");
publishOnStop = true;
stopRequested = true;
}
}
if (stopRequested) {
break;
}
// The retrying business is because the KafkaConsumer throws OffsetOutOfRangeException if the seeked-to
// offset is not present in the topic-partition. This can happen if we're asking a task to read from data
// that has not been written yet (which is totally legitimate). So let's wait for it to show up.
ConsumerRecords<byte[], byte[]> records = ConsumerRecords.empty();
try {
records = consumer.poll(POLL_TIMEOUT);
} catch (OffsetOutOfRangeException e) {
log.warn("OffsetOutOfRangeException with message [%s]", e.getMessage());
possiblyResetOffsetsOrWait(e.offsetOutOfRangePartitions(), consumer, toolbox);
stillReading = ioConfig.isPauseAfterRead() || !assignment.isEmpty();
}
for (ConsumerRecord<byte[], byte[]> record : records) {
if (log.isTraceEnabled()) {
log.trace("Got topic[%s] partition[%d] offset[%,d].", record.topic(), record.partition(), record.offset());
}
if (record.offset() < endOffsets.get(record.partition())) {
if (record.offset() != nextOffsets.get(record.partition())) {
throw new ISE("WTF?! Got offset[%,d] after offset[%,d] in partition[%d].", record.offset(), nextOffsets.get(record.partition()), record.partition());
}
try {
final byte[] valueBytes = record.value();
if (valueBytes == null) {
throw new ParseException("null value");
}
final InputRow row = Preconditions.checkNotNull(parser.parse(ByteBuffer.wrap(valueBytes)), "row");
if (!ioConfig.getMinimumMessageTime().isPresent() || !ioConfig.getMinimumMessageTime().get().isAfter(row.getTimestamp())) {
final SegmentIdentifier identifier = driver.add(row, sequenceNames.get(record.partition()), committerSupplier);
if (identifier == null) {
// If we allow continuing, then consider blacklisting the interval for a while to avoid constant checks.
throw new ISE("Could not allocate segment for row with timestamp[%s]", row.getTimestamp());
}
fireDepartmentMetrics.incrementProcessed();
} else {
fireDepartmentMetrics.incrementThrownAway();
}
} catch (ParseException e) {
if (tuningConfig.isReportParseExceptions()) {
throw e;
} else {
log.debug(e, "Dropping unparseable row from partition[%d] offset[%,d].", record.partition(), record.offset());
fireDepartmentMetrics.incrementUnparseable();
}
}
nextOffsets.put(record.partition(), record.offset() + 1);
}
if (nextOffsets.get(record.partition()).equals(endOffsets.get(record.partition())) && assignment.remove(record.partition())) {
log.info("Finished reading topic[%s], partition[%,d].", record.topic(), record.partition());
assignPartitions(consumer, topic, assignment);
stillReading = ioConfig.isPauseAfterRead() || !assignment.isEmpty();
}
}
}
} finally {
// persist pending data
driver.persist(committerSupplier.get());
}
synchronized (statusLock) {
if (stopRequested && !publishOnStop) {
throw new InterruptedException("Stopping without publishing");
}
status = Status.PUBLISHING;
}
final TransactionalSegmentPublisher publisher = new TransactionalSegmentPublisher() {
@Override
public boolean publishSegments(Set<DataSegment> segments, Object commitMetadata) throws IOException {
final KafkaPartitions finalPartitions = toolbox.getObjectMapper().convertValue(((Map) commitMetadata).get(METADATA_NEXT_PARTITIONS), KafkaPartitions.class);
// Sanity check, we should only be publishing things that match our desired end state.
if (!endOffsets.equals(finalPartitions.getPartitionOffsetMap())) {
throw new ISE("WTF?! Driver attempted to publish invalid metadata[%s].", commitMetadata);
}
final SegmentTransactionalInsertAction action;
if (ioConfig.isUseTransaction()) {
action = new SegmentTransactionalInsertAction(segments, new KafkaDataSourceMetadata(ioConfig.getStartPartitions()), new KafkaDataSourceMetadata(finalPartitions));
} else {
action = new SegmentTransactionalInsertAction(segments, null, null);
}
log.info("Publishing with isTransaction[%s].", ioConfig.isUseTransaction());
return toolbox.getTaskActionClient().submit(action).isSuccess();
}
};
final SegmentsAndMetadata published = driver.finish(publisher, committerSupplier.get());
if (published == null) {
throw new ISE("Transaction failure publishing segments, aborting");
} else {
log.info("Published segments[%s] with metadata[%s].", Joiner.on(", ").join(Iterables.transform(published.getSegments(), new Function<DataSegment, String>() {
@Override
public String apply(DataSegment input) {
return input.getIdentifier();
}
})), published.getCommitMetadata());
}
} catch (InterruptedException | RejectedExecutionException e) {
// handle the InterruptedException that gets wrapped in a RejectedExecutionException
if (e instanceof RejectedExecutionException && (e.getCause() == null || !(e.getCause() instanceof InterruptedException))) {
throw e;
}
// if we were interrupted because we were asked to stop, handle the exception and return success, else rethrow
if (!stopRequested) {
Thread.currentThread().interrupt();
throw e;
}
log.info("The task was asked to stop before completing");
} finally {
if (chatHandlerProvider.isPresent()) {
chatHandlerProvider.get().unregister(getId());
}
}
return success();
}
use of com.google.common.base.Supplier in project druid by druid-io.
the class GroupByRowProcessor method process.
public static Sequence<Row> process(final Query queryParam, final Sequence<Row> rows, final Map<String, ValueType> rowSignature, final GroupByQueryConfig config, final GroupByQueryResource resource, final ObjectMapper spillMapper, final String processingTmpDir) {
final GroupByQuery query = (GroupByQuery) queryParam;
final GroupByQueryConfig querySpecificConfig = config.withOverrides(query);
final AggregatorFactory[] aggregatorFactories = new AggregatorFactory[query.getAggregatorSpecs().size()];
for (int i = 0; i < query.getAggregatorSpecs().size(); i++) {
aggregatorFactories[i] = query.getAggregatorSpecs().get(i);
}
final File temporaryStorageDirectory = new File(processingTmpDir, String.format("druid-groupBy-%s_%s", UUID.randomUUID(), query.getId()));
final List<Interval> queryIntervals = query.getIntervals();
final Filter filter = Filters.convertToCNFFromQueryContext(query, Filters.toFilter(query.getDimFilter()));
final SettableSupplier<Row> rowSupplier = new SettableSupplier<>();
final RowBasedColumnSelectorFactory columnSelectorFactory = RowBasedColumnSelectorFactory.create(rowSupplier, rowSignature);
final ValueMatcher filterMatcher = filter == null ? BooleanValueMatcher.of(true) : filter.makeMatcher(columnSelectorFactory);
final FilteredSequence<Row> filteredSequence = new FilteredSequence<>(rows, new Predicate<Row>() {
@Override
public boolean apply(Row input) {
boolean inInterval = false;
DateTime rowTime = input.getTimestamp();
for (Interval queryInterval : queryIntervals) {
if (queryInterval.contains(rowTime)) {
inInterval = true;
break;
}
}
if (!inInterval) {
return false;
}
rowSupplier.set(input);
return filterMatcher.matches();
}
});
return new BaseSequence<>(new BaseSequence.IteratorMaker<Row, CloseableGrouperIterator<RowBasedKey, Row>>() {
@Override
public CloseableGrouperIterator<RowBasedKey, Row> make() {
// This contains all closeable objects which are closed when the returned iterator iterates all the elements,
// or an exceptions is thrown. The objects are closed in their reverse order.
final List<Closeable> closeOnExit = Lists.newArrayList();
try {
final LimitedTemporaryStorage temporaryStorage = new LimitedTemporaryStorage(temporaryStorageDirectory, querySpecificConfig.getMaxOnDiskStorage());
closeOnExit.add(temporaryStorage);
Pair<Grouper<RowBasedKey>, Accumulator<Grouper<RowBasedKey>, Row>> pair = RowBasedGrouperHelper.createGrouperAccumulatorPair(query, true, rowSignature, querySpecificConfig, new Supplier<ByteBuffer>() {
@Override
public ByteBuffer get() {
final ResourceHolder<ByteBuffer> mergeBufferHolder = resource.getMergeBuffer();
closeOnExit.add(mergeBufferHolder);
return mergeBufferHolder.get();
}
}, -1, temporaryStorage, spillMapper, aggregatorFactories);
final Grouper<RowBasedKey> grouper = pair.lhs;
final Accumulator<Grouper<RowBasedKey>, Row> accumulator = pair.rhs;
closeOnExit.add(grouper);
final Grouper<RowBasedKey> retVal = filteredSequence.accumulate(grouper, accumulator);
if (retVal != grouper) {
throw GroupByQueryHelper.throwAccumulationResourceLimitExceededException();
}
return RowBasedGrouperHelper.makeGrouperIterator(grouper, query, new Closeable() {
@Override
public void close() throws IOException {
for (Closeable closeable : Lists.reverse(closeOnExit)) {
CloseQuietly.close(closeable);
}
}
});
} catch (Throwable e) {
// Exception caught while setting up the iterator; release resources.
for (Closeable closeable : Lists.reverse(closeOnExit)) {
CloseQuietly.close(closeable);
}
throw e;
}
}
@Override
public void cleanup(CloseableGrouperIterator<RowBasedKey, Row> iterFromMake) {
iterFromMake.close();
}
});
}
Aggregations