use of org.apache.druid.indexer.TaskStatus in project druid by druid-io.
the class CompactionTaskRunTest method testPartialIntervalCompactWithFinerSegmentGranularityThenFullIntervalCompactWithDropExistingTrue.
@Test
public void testPartialIntervalCompactWithFinerSegmentGranularityThenFullIntervalCompactWithDropExistingTrue() throws Exception {
// This test fails with segment lock because of the bug reported in https://github.com/apache/druid/issues/10911.
if (lockGranularity == LockGranularity.SEGMENT) {
return;
}
// This creates HOUR segments with intervals of
// - 2014-01-01T00:00:00/2014-01-01T01:00:00
// - 2014-01-01T01:00:00/2014-01-01T02:00:00
// - 2014-01-01T02:00:00/2014-01-01T03:00:00
runIndexTask();
final Interval compactionPartialInterval = Intervals.of("2014-01-01T01:00:00/2014-01-01T02:00:00");
// Segments that did not belong in the compaction interval are expected unchanged
final Set<DataSegment> expectedSegments = new HashSet<>();
expectedSegments.addAll(getStorageCoordinator().retrieveUsedSegmentsForIntervals(DATA_SOURCE, Collections.singletonList(Intervals.of("2014-01-01T02:00:00/2014-01-01T03:00:00")), Segments.ONLY_VISIBLE));
expectedSegments.addAll(getStorageCoordinator().retrieveUsedSegmentsForIntervals(DATA_SOURCE, Collections.singletonList(Intervals.of("2014-01-01T00:00:00/2014-01-01T01:00:00")), Segments.ONLY_VISIBLE));
final Builder builder = new Builder(DATA_SOURCE, segmentCacheManagerFactory, RETRY_POLICY_FACTORY);
final CompactionTask partialCompactionTask = builder.segmentGranularity(Granularities.MINUTE).inputSpec(new CompactionIntervalSpec(compactionPartialInterval, null), true).build();
final Pair<TaskStatus, List<DataSegment>> partialCompactionResult = runTask(partialCompactionTask);
Assert.assertTrue(partialCompactionResult.lhs.isSuccess());
// New segments that was compacted are expected. However, old segments of the compacted interval should be drop
// regardless of the new segments fully overshadow the old segments or not. Hence, we do not expect old segments
// of the 2014-01-01T01:00:00/2014-01-01T02:00:00 interval post-compaction
expectedSegments.addAll(partialCompactionResult.rhs);
final Set<DataSegment> segmentsAfterPartialCompaction = new HashSet<>(getStorageCoordinator().retrieveUsedSegmentsForIntervals(DATA_SOURCE, Collections.singletonList(Intervals.of("2014-01-01/2014-01-02")), Segments.ONLY_VISIBLE));
Assert.assertEquals(expectedSegments, segmentsAfterPartialCompaction);
final CompactionTask fullCompactionTask = builder.segmentGranularity(null).inputSpec(new CompactionIntervalSpec(Intervals.of("2014-01-01/2014-01-02"), null), true).build();
final Pair<TaskStatus, List<DataSegment>> fullCompactionResult = runTask(fullCompactionTask);
Assert.assertTrue(fullCompactionResult.lhs.isSuccess());
final List<DataSegment> segmentsAfterFullCompaction = new ArrayList<>(getStorageCoordinator().retrieveUsedSegmentsForIntervals(DATA_SOURCE, Collections.singletonList(Intervals.of("2014-01-01/2014-01-02")), Segments.ONLY_VISIBLE));
segmentsAfterFullCompaction.sort((s1, s2) -> Comparators.intervalsByStartThenEnd().compare(s1.getInterval(), s2.getInterval()));
Assert.assertEquals(3, segmentsAfterFullCompaction.size());
// Full Compaction with null segmentGranularity meaning that the original segmentGrnaularity is perserved
// For the intervals, 2014-01-01T00:00:00.000Z/2014-01-01T01:00:00.000Z and 2014-01-01T02:00:00.000Z/2014-01-01T03:00:00.000Z
// the original segmentGranularity is HOUR from the initial ingestion.
// For the interval, 2014-01-01T01:00:00.000Z/2014-01-01T01:01:00.000Z, the original segmentGranularity is
// MINUTE from the partial compaction done earlier.
Assert.assertEquals(Intervals.of("2014-01-01T00:00:00.000Z/2014-01-01T01:00:00.000Z"), segmentsAfterFullCompaction.get(0).getInterval());
Assert.assertEquals(Intervals.of("2014-01-01T01:00:00.000Z/2014-01-01T01:01:00.000Z"), segmentsAfterFullCompaction.get(1).getInterval());
Assert.assertEquals(Intervals.of("2014-01-01T02:00:00.000Z/2014-01-01T03:00:00.000Z"), segmentsAfterFullCompaction.get(2).getInterval());
}
use of org.apache.druid.indexer.TaskStatus in project druid by druid-io.
the class CompactionTaskRunTest method testRunIndexAndCompactForSameSegmentAtTheSameTime.
@Test
public void testRunIndexAndCompactForSameSegmentAtTheSameTime() throws Exception {
runIndexTask();
// make sure that indexTask becomes ready first, then compactionTask becomes ready, then indexTask runs
final CountDownLatch compactionTaskReadyLatch = new CountDownLatch(1);
final CountDownLatch indexTaskStartLatch = new CountDownLatch(1);
final Future<Pair<TaskStatus, List<DataSegment>>> indexFuture = exec.submit(() -> runIndexTask(compactionTaskReadyLatch, indexTaskStartLatch, false));
final Builder builder = new Builder(DATA_SOURCE, segmentCacheManagerFactory, RETRY_POLICY_FACTORY);
final CompactionTask compactionTask = builder.interval(Intervals.of("2014-01-01T00:00:00/2014-01-02T03:00:00")).build();
final Future<Pair<TaskStatus, List<DataSegment>>> compactionFuture = exec.submit(() -> {
compactionTaskReadyLatch.await();
return runTask(compactionTask, indexTaskStartLatch, null);
});
Assert.assertTrue(indexFuture.get().lhs.isSuccess());
List<DataSegment> segments = indexFuture.get().rhs;
Assert.assertEquals(6, segments.size());
for (int i = 0; i < 6; i++) {
Assert.assertEquals(Intervals.of("2014-01-01T0%d:00:00/2014-01-01T0%d:00:00", i / 2, i / 2 + 1), segments.get(i).getInterval());
if (lockGranularity == LockGranularity.SEGMENT) {
Assert.assertEquals(new NumberedOverwriteShardSpec(PartitionIds.NON_ROOT_GEN_START_PARTITION_ID + i % 2, 0, 2, (short) 1, (short) 2), segments.get(i).getShardSpec());
} else {
Assert.assertEquals(new NumberedShardSpec(i % 2, 2), segments.get(i).getShardSpec());
}
}
final Pair<TaskStatus, List<DataSegment>> compactionResult = compactionFuture.get();
Assert.assertEquals(TaskState.FAILED, compactionResult.lhs.getStatusCode());
}
use of org.apache.druid.indexer.TaskStatus in project druid by druid-io.
the class CompactionTaskRunTest method testPartialIntervalCompactWithFinerSegmentGranularityThenFullIntervalCompactWithDropExistingFalse.
@Test
public void testPartialIntervalCompactWithFinerSegmentGranularityThenFullIntervalCompactWithDropExistingFalse() throws Exception {
// This test fails with segment lock because of the bug reported in https://github.com/apache/druid/issues/10911.
if (lockGranularity == LockGranularity.SEGMENT) {
return;
}
runIndexTask();
final Set<DataSegment> expectedSegments = new HashSet<>(getStorageCoordinator().retrieveUsedSegmentsForIntervals(DATA_SOURCE, Collections.singletonList(Intervals.of("2014-01-01/2014-01-02")), Segments.ONLY_VISIBLE));
final Builder builder = new Builder(DATA_SOURCE, segmentCacheManagerFactory, RETRY_POLICY_FACTORY);
final Interval partialInterval = Intervals.of("2014-01-01T01:00:00/2014-01-01T02:00:00");
final CompactionTask partialCompactionTask = builder.segmentGranularity(Granularities.MINUTE).inputSpec(new CompactionIntervalSpec(partialInterval, null), false).build();
final Pair<TaskStatus, List<DataSegment>> partialCompactionResult = runTask(partialCompactionTask);
Assert.assertTrue(partialCompactionResult.lhs.isSuccess());
// All segments in the previous expectedSegments should still appear as they have larger segment granularity.
expectedSegments.addAll(partialCompactionResult.rhs);
final Set<DataSegment> segmentsAfterPartialCompaction = new HashSet<>(getStorageCoordinator().retrieveUsedSegmentsForIntervals(DATA_SOURCE, Collections.singletonList(Intervals.of("2014-01-01/2014-01-02")), Segments.ONLY_VISIBLE));
Assert.assertEquals(expectedSegments, segmentsAfterPartialCompaction);
final CompactionTask fullCompactionTask = builder.segmentGranularity(null).inputSpec(new CompactionIntervalSpec(Intervals.of("2014-01-01/2014-01-02"), null), false).build();
final Pair<TaskStatus, List<DataSegment>> fullCompactionResult = runTask(fullCompactionTask);
Assert.assertTrue(fullCompactionResult.lhs.isSuccess());
final List<DataSegment> segmentsAfterFullCompaction = new ArrayList<>(getStorageCoordinator().retrieveUsedSegmentsForIntervals(DATA_SOURCE, Collections.singletonList(Intervals.of("2014-01-01/2014-01-02")), Segments.ONLY_VISIBLE));
segmentsAfterFullCompaction.sort((s1, s2) -> Comparators.intervalsByStartThenEnd().compare(s1.getInterval(), s2.getInterval()));
Assert.assertEquals(3, segmentsAfterFullCompaction.size());
for (int i = 0; i < segmentsAfterFullCompaction.size(); i++) {
Assert.assertEquals(Intervals.of(StringUtils.format("2014-01-01T%02d/2014-01-01T%02d", i, i + 1)), segmentsAfterFullCompaction.get(i).getInterval());
}
}
use of org.apache.druid.indexer.TaskStatus in project druid by druid-io.
the class AppenderatorDriverRealtimeIndexTaskTest method testBasics.
@Test(timeout = 60_000L)
public void testBasics() throws Exception {
expectPublishedSegments(1);
final AppenderatorDriverRealtimeIndexTask task = makeRealtimeTask(null);
Assert.assertTrue(task.supportsQueries());
final ListenableFuture<TaskStatus> statusFuture = runTask(task);
// Wait for firehose to show up, it starts off null.
while (task.getFirehose() == null) {
Thread.sleep(50);
}
final TestFirehose firehose = (TestFirehose) task.getFirehose();
firehose.addRows(ImmutableList.of(ImmutableMap.of("t", now.getMillis(), "dim1", "foo", "met1", "1"), ImmutableMap.of("t", now.getMillis(), "dim2", "bar", "met1", 2.0)));
// Stop the firehose, this will drain out existing events.
firehose.close();
// Wait for publish.
Collection<DataSegment> publishedSegments = awaitSegments();
// Check metrics.
Assert.assertEquals(2, task.getRowIngestionMeters().getProcessed());
Assert.assertEquals(0, task.getRowIngestionMeters().getThrownAway());
Assert.assertEquals(0, task.getRowIngestionMeters().getUnparseable());
// Do some queries.
Assert.assertEquals(2, sumMetric(task, null, "rows").longValue());
Assert.assertEquals(3, sumMetric(task, null, "met1").longValue());
awaitHandoffs();
for (DataSegment publishedSegment : publishedSegments) {
Pair<Executor, Runnable> executorRunnablePair = handOffCallbacks.get(new SegmentDescriptor(publishedSegment.getInterval(), publishedSegment.getVersion(), publishedSegment.getShardSpec().getPartitionNum()));
Assert.assertNotNull(publishedSegment + " missing from handoff callbacks: " + handOffCallbacks, executorRunnablePair);
// Simulate handoff.
executorRunnablePair.lhs.execute(executorRunnablePair.rhs);
}
handOffCallbacks.clear();
// Wait for the task to finish.
final TaskStatus taskStatus = statusFuture.get();
Assert.assertEquals(TaskState.SUCCESS, taskStatus.getStatusCode());
}
use of org.apache.druid.indexer.TaskStatus in project druid by druid-io.
the class AppenderatorDriverRealtimeIndexTaskTest method testTransformSpec.
@Test(timeout = 60_000L)
public void testTransformSpec() throws Exception {
expectPublishedSegments(2);
final TransformSpec transformSpec = new TransformSpec(new SelectorDimFilter("dim1", "foo", null), ImmutableList.of(new ExpressionTransform("dim1t", "concat(dim1,dim1)", ExprMacroTable.nil())));
final AppenderatorDriverRealtimeIndexTask task = makeRealtimeTask(null, transformSpec, true, 0, true, 0, 1);
final ListenableFuture<TaskStatus> statusFuture = runTask(task);
// Wait for firehose to show up, it starts off null.
while (task.getFirehose() == null) {
Thread.sleep(50);
}
final TestFirehose firehose = (TestFirehose) task.getFirehose();
firehose.addRows(ImmutableList.of(ImmutableMap.of("t", now.getMillis(), "dim1", "foo", "met1", "1"), ImmutableMap.of("t", now.minus(new Period("P1D")).getMillis(), "dim1", "foo", "met1", 2.0), ImmutableMap.of("t", now.getMillis(), "dim2", "bar", "met1", 2.0)));
// Stop the firehose, this will drain out existing events.
firehose.close();
Collection<DataSegment> publishedSegments = awaitSegments();
// Check metrics.
Assert.assertEquals(2, task.getRowIngestionMeters().getProcessed());
Assert.assertEquals(1, task.getRowIngestionMeters().getThrownAway());
Assert.assertEquals(0, task.getRowIngestionMeters().getUnparseable());
// Do some queries.
Assert.assertEquals(2, sumMetric(task, null, "rows").longValue());
Assert.assertEquals(2, sumMetric(task, new SelectorDimFilter("dim1t", "foofoo", null), "rows").longValue());
if (NullHandling.replaceWithDefault()) {
Assert.assertEquals(0, sumMetric(task, new SelectorDimFilter("dim1t", "barbar", null), "metric1").longValue());
} else {
Assert.assertNull(sumMetric(task, new SelectorDimFilter("dim1t", "barbar", null), "metric1"));
}
Assert.assertEquals(3, sumMetric(task, null, "met1").longValue());
awaitHandoffs();
for (DataSegment publishedSegment : publishedSegments) {
Pair<Executor, Runnable> executorRunnablePair = handOffCallbacks.get(new SegmentDescriptor(publishedSegment.getInterval(), publishedSegment.getVersion(), publishedSegment.getShardSpec().getPartitionNum()));
Assert.assertNotNull(publishedSegment + " missing from handoff callbacks: " + handOffCallbacks, executorRunnablePair);
// Simulate handoff.
executorRunnablePair.lhs.execute(executorRunnablePair.rhs);
}
handOffCallbacks.clear();
// Wait for the task to finish.
final TaskStatus taskStatus = statusFuture.get();
Assert.assertEquals(TaskState.SUCCESS, taskStatus.getStatusCode());
}
Aggregations