use of org.apache.druid.client.selector.ServerSelector in project druid by druid-io.
the class ClientInfoResource method getDatasource.
@GET
@Path("/{dataSourceName}")
@Produces(MediaType.APPLICATION_JSON)
@ResourceFilters(DatasourceResourceFilter.class)
public Map<String, Object> getDatasource(@PathParam("dataSourceName") String dataSourceName, @QueryParam("interval") String interval, @QueryParam("full") String full) {
if (full == null) {
return ImmutableMap.of(KEY_DIMENSIONS, getDataSourceDimensions(dataSourceName, interval), KEY_METRICS, getDataSourceMetrics(dataSourceName, interval));
}
Interval theInterval;
if (interval == null || interval.isEmpty()) {
DateTime now = getCurrentTime();
theInterval = new Interval(segmentMetadataQueryConfig.getDefaultHistory(), now);
} else {
theInterval = Intervals.of(interval);
}
final Optional<? extends TimelineLookup<String, ServerSelector>> maybeTimeline = timelineServerView.getTimeline(DataSourceAnalysis.forDataSource(new TableDataSource(dataSourceName)));
final Optional<Iterable<TimelineObjectHolder<String, ServerSelector>>> maybeServersLookup = maybeTimeline.map(timeline -> timeline.lookup(theInterval));
if (!maybeServersLookup.isPresent() || Iterables.isEmpty(maybeServersLookup.get())) {
return Collections.emptyMap();
}
Map<Interval, Object> servedIntervals = new TreeMap<>(new Comparator<Interval>() {
@Override
public int compare(Interval o1, Interval o2) {
if (o1.equals(o2) || o1.overlaps(o2)) {
return 0;
} else {
return o1.isBefore(o2) ? -1 : 1;
}
}
});
for (TimelineObjectHolder<String, ServerSelector> holder : maybeServersLookup.get()) {
final Set<Object> dimensions = new HashSet<>();
final Set<Object> metrics = new HashSet<>();
final PartitionHolder<ServerSelector> partitionHolder = holder.getObject();
if (partitionHolder.isComplete()) {
for (ServerSelector server : partitionHolder.payloads()) {
final DataSegment segment = server.getSegment();
dimensions.addAll(segment.getDimensions());
metrics.addAll(segment.getMetrics());
}
}
servedIntervals.put(holder.getInterval(), ImmutableMap.of(KEY_DIMENSIONS, dimensions, KEY_METRICS, metrics));
}
// collapse intervals if they abut and have same set of columns
Map<String, Object> result = Maps.newLinkedHashMap();
Interval curr = null;
Map<String, Set<String>> cols = null;
for (Map.Entry<Interval, Object> e : servedIntervals.entrySet()) {
Interval ival = e.getKey();
if (curr != null && curr.abuts(ival) && cols.equals(e.getValue())) {
curr = curr.withEnd(ival.getEnd());
} else {
if (curr != null) {
result.put(curr.toString(), cols);
}
curr = ival;
cols = (Map<String, Set<String>>) e.getValue();
}
}
// add the last one in
if (curr != null) {
result.put(curr.toString(), cols);
}
return result;
}
use of org.apache.druid.client.selector.ServerSelector in project druid by druid-io.
the class CachingClusteredClientTest method testEtagforDifferentQueryInterval.
@Test
public void testEtagforDifferentQueryInterval() {
final Interval interval = Intervals.of("2016-01-01/2016-01-02");
final Interval queryInterval = Intervals.of("2016-01-01T14:00:00/2016-01-02T14:00:00");
final Interval queryInterval2 = Intervals.of("2016-01-01T18:00:00/2016-01-02T18:00:00");
final DataSegment dataSegment = new DataSegment("dataSource", interval, "ver", ImmutableMap.of("type", "hdfs", "path", "/tmp"), ImmutableList.of("product"), ImmutableList.of("visited_sum"), NoneShardSpec.instance(), 9, 12334);
final ServerSelector selector = new ServerSelector(dataSegment, new HighestPriorityTierSelectorStrategy(new RandomServerSelectorStrategy()));
selector.addServerAndUpdateSegment(new QueryableDruidServer(servers[0], null), dataSegment);
timeline.add(interval, "ver", new SingleElementPartitionChunk<>(selector));
final TimeBoundaryQuery query = Druids.newTimeBoundaryQueryBuilder().dataSource(DATA_SOURCE).intervals(new MultipleIntervalSegmentSpec(ImmutableList.of(queryInterval))).context(ImmutableMap.of("If-None-Match", "aVJV29CJY93rszVW/QBy0arWZo0=")).randomQueryId().build();
final TimeBoundaryQuery query2 = Druids.newTimeBoundaryQueryBuilder().dataSource(DATA_SOURCE).intervals(new MultipleIntervalSegmentSpec(ImmutableList.of(queryInterval2))).context(ImmutableMap.of("If-None-Match", "aVJV29CJY93rszVW/QBy0arWZo0=")).randomQueryId().build();
final ResponseContext responseContext = initializeResponseContext();
getDefaultQueryRunner().run(QueryPlus.wrap(query), responseContext);
final String etag1 = responseContext.getEntityTag();
getDefaultQueryRunner().run(QueryPlus.wrap(query2), responseContext);
final String etag2 = responseContext.getEntityTag();
Assert.assertNotEquals(etag1, etag2);
}
use of org.apache.druid.client.selector.ServerSelector in project druid by druid-io.
the class CachingClusteredClientTest method testCachingOverBulkLimitEnforcesLimit.
@Test
@SuppressWarnings("unchecked")
public void testCachingOverBulkLimitEnforcesLimit() {
final int limit = 10;
final Interval interval = Intervals.of("2011-01-01/2011-01-02");
final TimeseriesQuery query = Druids.newTimeseriesQueryBuilder().dataSource(DATA_SOURCE).intervals(new MultipleIntervalSegmentSpec(ImmutableList.of(interval))).filters(DIM_FILTER).granularity(GRANULARITY).aggregators(AGGS).postAggregators(POST_AGGS).context(CONTEXT).randomQueryId().build();
final ResponseContext context = initializeResponseContext();
final Cache cache = EasyMock.createStrictMock(Cache.class);
final Capture<Iterable<Cache.NamedKey>> cacheKeyCapture = EasyMock.newCapture();
EasyMock.expect(cache.getBulk(EasyMock.capture(cacheKeyCapture))).andReturn(ImmutableMap.of()).once();
EasyMock.replay(cache);
client = makeClient(new ForegroundCachePopulator(JSON_MAPPER, new CachePopulatorStats(), -1), cache, limit);
final DruidServer lastServer = servers[random.nextInt(servers.length)];
final DataSegment dataSegment = EasyMock.createNiceMock(DataSegment.class);
EasyMock.expect(dataSegment.getId()).andReturn(SegmentId.dummy(DATA_SOURCE)).anyTimes();
EasyMock.replay(dataSegment);
final ServerSelector selector = new ServerSelector(dataSegment, new HighestPriorityTierSelectorStrategy(new RandomServerSelectorStrategy()));
selector.addServerAndUpdateSegment(new QueryableDruidServer(lastServer, null), dataSegment);
timeline.add(interval, "v", new SingleElementPartitionChunk<>(selector));
getDefaultQueryRunner().run(QueryPlus.wrap(query), context);
Assert.assertTrue("Capture cache keys", cacheKeyCapture.hasCaptured());
Assert.assertTrue("Cache key below limit", ImmutableList.copyOf(cacheKeyCapture.getValue()).size() <= limit);
EasyMock.verify(cache);
EasyMock.reset(cache);
cacheKeyCapture.reset();
EasyMock.expect(cache.getBulk(EasyMock.capture(cacheKeyCapture))).andReturn(ImmutableMap.of()).once();
EasyMock.replay(cache);
client = makeClient(new ForegroundCachePopulator(JSON_MAPPER, new CachePopulatorStats(), -1), cache, 0);
getDefaultQueryRunner().run(QueryPlus.wrap(query), context);
EasyMock.verify(cache);
EasyMock.verify(dataSegment);
Assert.assertTrue("Capture cache keys", cacheKeyCapture.hasCaptured());
Assert.assertTrue("Cache Keys empty", ImmutableList.copyOf(cacheKeyCapture.getValue()).isEmpty());
}
use of org.apache.druid.client.selector.ServerSelector in project druid by druid-io.
the class CachingClusteredClientTest method testSingleDimensionPruning.
@Test
public void testSingleDimensionPruning() {
DimFilter filter = new AndDimFilter(new OrDimFilter(new SelectorDimFilter("dim1", "a", null), new BoundDimFilter("dim1", "from", "to", false, false, false, null, StringComparators.LEXICOGRAPHIC)), new AndDimFilter(new InDimFilter("dim2", Arrays.asList("a", "c", "e", "g"), null), new BoundDimFilter("dim2", "aaa", "hi", false, false, false, null, StringComparators.LEXICOGRAPHIC), new BoundDimFilter("dim2", "e", "zzz", true, true, false, null, StringComparators.LEXICOGRAPHIC)));
final Druids.TimeseriesQueryBuilder builder = Druids.newTimeseriesQueryBuilder().dataSource(DATA_SOURCE).filters(filter).granularity(GRANULARITY).intervals(SEG_SPEC).context(CONTEXT).intervals("2011-01-05/2011-01-10").aggregators(RENAMED_AGGS).postAggregators(RENAMED_POST_AGGS);
TimeseriesQuery query = builder.randomQueryId().build();
final Interval interval1 = Intervals.of("2011-01-06/2011-01-07");
final Interval interval2 = Intervals.of("2011-01-07/2011-01-08");
final Interval interval3 = Intervals.of("2011-01-08/2011-01-09");
QueryRunner runner = new FinalizeResultsQueryRunner(getDefaultQueryRunner(), new TimeseriesQueryQueryToolChest());
final DruidServer lastServer = servers[random.nextInt(servers.length)];
ServerSelector selector1 = makeMockSingleDimensionSelector(lastServer, "dim1", null, "b", 0);
ServerSelector selector2 = makeMockSingleDimensionSelector(lastServer, "dim1", "e", "f", 1);
ServerSelector selector3 = makeMockSingleDimensionSelector(lastServer, "dim1", "hi", "zzz", 2);
ServerSelector selector4 = makeMockSingleDimensionSelector(lastServer, "dim2", "a", "e", 0);
ServerSelector selector5 = makeMockSingleDimensionSelector(lastServer, "dim2", null, null, 1);
ServerSelector selector6 = makeMockSingleDimensionSelector(lastServer, "other", "b", null, 0);
timeline.add(interval1, "v", new NumberedPartitionChunk<>(0, 3, selector1));
timeline.add(interval1, "v", new NumberedPartitionChunk<>(1, 3, selector2));
timeline.add(interval1, "v", new NumberedPartitionChunk<>(2, 3, selector3));
timeline.add(interval2, "v", new NumberedPartitionChunk<>(0, 2, selector4));
timeline.add(interval2, "v", new NumberedPartitionChunk<>(1, 2, selector5));
timeline.add(interval3, "v", new NumberedPartitionChunk<>(0, 1, selector6));
final Capture<QueryPlus> capture = Capture.newInstance();
final Capture<ResponseContext> contextCap = Capture.newInstance();
QueryRunner mockRunner = EasyMock.createNiceMock(QueryRunner.class);
EasyMock.expect(mockRunner.run(EasyMock.capture(capture), EasyMock.capture(contextCap))).andReturn(Sequences.empty()).anyTimes();
EasyMock.expect(serverView.getQueryRunner(lastServer)).andReturn(mockRunner).anyTimes();
EasyMock.replay(serverView);
EasyMock.replay(mockRunner);
List<SegmentDescriptor> descriptors = new ArrayList<>();
descriptors.add(new SegmentDescriptor(interval1, "v", 0));
descriptors.add(new SegmentDescriptor(interval1, "v", 2));
descriptors.add(new SegmentDescriptor(interval2, "v", 1));
descriptors.add(new SegmentDescriptor(interval3, "v", 0));
MultipleSpecificSegmentSpec expected = new MultipleSpecificSegmentSpec(descriptors);
runner.run(QueryPlus.wrap(query)).toList();
Assert.assertEquals(expected, ((TimeseriesQuery) capture.getValue().getQuery()).getQuerySegmentSpec());
}
use of org.apache.druid.client.selector.ServerSelector in project druid by druid-io.
the class CachingClusteredClientPerfTest method testGetQueryRunnerForSegments_singleIntervalLargeSegments.
@Test(timeout = 10_000)
public void testGetQueryRunnerForSegments_singleIntervalLargeSegments() {
final int segmentCount = 30_000;
final Interval interval = Intervals.of("2021-02-13/2021-02-14");
final List<SegmentDescriptor> segmentDescriptors = new ArrayList<>(segmentCount);
final List<DataSegment> dataSegments = new ArrayList<>(segmentCount);
final VersionedIntervalTimeline<String, ServerSelector> timeline = new VersionedIntervalTimeline<>(Ordering.natural());
final DruidServer server = new DruidServer("server", "localhost:9000", null, Long.MAX_VALUE, ServerType.HISTORICAL, DruidServer.DEFAULT_TIER, DruidServer.DEFAULT_PRIORITY);
for (int ii = 0; ii < segmentCount; ii++) {
segmentDescriptors.add(new SegmentDescriptor(interval, "1", ii));
DataSegment segment = makeDataSegment("test", interval, "1", ii);
dataSegments.add(segment);
}
timeline.addAll(Iterators.transform(dataSegments.iterator(), segment -> {
ServerSelector ss = new ServerSelector(segment, new HighestPriorityTierSelectorStrategy(new RandomServerSelectorStrategy()));
ss.addServerAndUpdateSegment(new QueryableDruidServer(server, new MockQueryRunner()), segment);
return new VersionedIntervalTimeline.PartitionChunkEntry<>(segment.getInterval(), segment.getVersion(), segment.getShardSpec().createChunk(ss));
}));
TimelineServerView serverView = Mockito.mock(TimelineServerView.class);
QueryScheduler queryScheduler = Mockito.mock(QueryScheduler.class);
// mock scheduler to return same sequence as argument
Mockito.when(queryScheduler.run(any(), any())).thenAnswer(i -> i.getArgument(1));
Mockito.when(queryScheduler.prioritizeAndLaneQuery(any(), any())).thenAnswer(i -> ((QueryPlus) i.getArgument(0)).getQuery());
Mockito.doReturn(Optional.of(timeline)).when(serverView).getTimeline(any());
Mockito.doReturn(new MockQueryRunner()).when(serverView).getQueryRunner(any());
CachingClusteredClient cachingClusteredClient = new CachingClusteredClient(new MockQueryToolChestWareHouse(), serverView, MapCache.create(1024), TestHelper.makeJsonMapper(), Mockito.mock(CachePopulator.class), new CacheConfig(), Mockito.mock(DruidHttpClientConfig.class), Mockito.mock(DruidProcessingConfig.class), ForkJoinPool.commonPool(), queryScheduler, NoopJoinableFactory.INSTANCE, new NoopServiceEmitter());
Query<SegmentDescriptor> fakeQuery = makeFakeQuery(interval);
QueryRunner<SegmentDescriptor> queryRunner = cachingClusteredClient.getQueryRunnerForSegments(fakeQuery, segmentDescriptors);
Sequence<SegmentDescriptor> sequence = queryRunner.run(QueryPlus.wrap(fakeQuery));
Assert.assertEquals(segmentDescriptors, sequence.toList());
}
Aggregations