Examples with DataSegmentPusher org.apache.druid.segment.loading.DataSegmentPusher used on opensource projects

Example 6 with DataSegmentPusher

use of org.apache.druid.segment.loading.DataSegmentPusher in project hive by apache.

the class TestDruidStorageHandler method testCommitMultiInsertOverwriteTable.

@Test
public void testCommitMultiInsertOverwriteTable() throws MetaException, IOException {
    DerbyConnectorTestUtility connector = derbyConnectorRule.getConnector();
    MetadataStorageTablesConfig metadataStorageTablesConfig = derbyConnectorRule.metadataTablesConfigSupplier().get();
    LocalFileSystem localFileSystem = FileSystem.getLocal(config);
    druidStorageHandler.preCreateTable(tableMock);
    Path taskDirPath = new Path(tableWorkingPath, druidStorageHandler.makeStagingName());
    HdfsDataSegmentPusherConfig pusherConfig = new HdfsDataSegmentPusherConfig();
    pusherConfig.setStorageDirectory(config.get(String.valueOf(HiveConf.ConfVars.DRUID_SEGMENT_DIRECTORY)));
    DataSegmentPusher dataSegmentPusher = new HdfsDataSegmentPusher(pusherConfig, config, DruidStorageHandlerUtils.JSON_MAPPER);
    // This create and publish the segment to be overwritten
    List<DataSegment> existingSegments = Collections.singletonList(createSegment(new Path(taskDirPath, DruidStorageHandlerUtils.INDEX_ZIP).toString(), new Interval(100, 150, DateTimeZone.UTC), "v0", new LinearShardSpec(0)));
    DruidStorageHandlerUtils.publishSegmentsAndCommit(connector, metadataStorageTablesConfig, DATA_SOURCE_NAME, existingSegments, true, config, dataSegmentPusher);
    // Check that there is one datasource with the published segment
    Assert.assertArrayEquals(Lists.newArrayList(DATA_SOURCE_NAME).toArray(), Lists.newArrayList(DruidStorageHandlerUtils.getAllDataSourceNames(connector, metadataStorageTablesConfig)).toArray());
    // Sequence is the following:
    // 1) INSERT with no segments -> Original segment still present in the datasource
    // 2) INSERT OVERWRITE with no segments -> Datasource is empty
    // 3) INSERT OVERWRITE with no segments -> Datasource is empty
    // 4) INSERT with no segments -> Datasource is empty
    // 5) INSERT with one segment -> Datasource has one segment
    // 6) INSERT OVERWRITE with one segment -> Datasource has one segment
    // 7) INSERT with one segment -> Datasource has two segments
    // 8) INSERT OVERWRITE with no segments -> Datasource is empty
    // We start:
    // #1
    druidStorageHandler.commitInsertTable(tableMock, false);
    Assert.assertArrayEquals(Lists.newArrayList(DATA_SOURCE_NAME).toArray(), Lists.newArrayList(DruidStorageHandlerUtils.getAllDataSourceNames(connector, metadataStorageTablesConfig)).toArray());
    Assert.assertEquals(1, getUsedSegmentsList(connector, metadataStorageTablesConfig).size());
    // #2
    druidStorageHandler.commitInsertTable(tableMock, true);
    Assert.assertEquals(0, getUsedSegmentsList(connector, metadataStorageTablesConfig).size());
    // #3
    druidStorageHandler.commitInsertTable(tableMock, true);
    Assert.assertEquals(0, getUsedSegmentsList(connector, metadataStorageTablesConfig).size());
    // #4
    druidStorageHandler.commitInsertTable(tableMock, true);
    Assert.assertEquals(0, getUsedSegmentsList(connector, metadataStorageTablesConfig).size());
    // #5
    DataSegment dataSegment1 = createSegment(new Path(taskDirPath, DruidStorageHandlerUtils.INDEX_ZIP).toString(), new Interval(180, 250, DateTimeZone.UTC), "v1", new LinearShardSpec(0));
    Path descriptorPath1 = DruidStorageHandlerUtils.makeSegmentDescriptorOutputPath(dataSegment1, new Path(taskDirPath, DruidStorageHandler.SEGMENTS_DESCRIPTOR_DIR_NAME));
    DruidStorageHandlerUtils.writeSegmentDescriptor(localFileSystem, dataSegment1, descriptorPath1);
    druidStorageHandler.commitInsertTable(tableMock, false);
    Assert.assertArrayEquals(Lists.newArrayList(DATA_SOURCE_NAME).toArray(), Lists.newArrayList(DruidStorageHandlerUtils.getAllDataSourceNames(connector, metadataStorageTablesConfig)).toArray());
    Assert.assertEquals(1, getUsedSegmentsList(connector, metadataStorageTablesConfig).size());
    // #6
    DataSegment dataSegment2 = createSegment(new Path(taskDirPath, DruidStorageHandlerUtils.INDEX_ZIP).toString(), new Interval(200, 250, DateTimeZone.UTC), "v1", new LinearShardSpec(0));
    Path descriptorPath2 = DruidStorageHandlerUtils.makeSegmentDescriptorOutputPath(dataSegment2, new Path(taskDirPath, DruidStorageHandler.SEGMENTS_DESCRIPTOR_DIR_NAME));
    DruidStorageHandlerUtils.writeSegmentDescriptor(localFileSystem, dataSegment2, descriptorPath2);
    druidStorageHandler.commitInsertTable(tableMock, true);
    Assert.assertArrayEquals(Lists.newArrayList(DATA_SOURCE_NAME).toArray(), Lists.newArrayList(DruidStorageHandlerUtils.getAllDataSourceNames(connector, metadataStorageTablesConfig)).toArray());
    Assert.assertEquals(1, getUsedSegmentsList(connector, metadataStorageTablesConfig).size());
    // #7
    DataSegment dataSegment3 = createSegment(new Path(taskDirPath, DruidStorageHandlerUtils.INDEX_ZIP).toString(), new Interval(100, 200, DateTimeZone.UTC), "v1", new LinearShardSpec(0));
    Path descriptorPath3 = DruidStorageHandlerUtils.makeSegmentDescriptorOutputPath(dataSegment3, new Path(taskDirPath, DruidStorageHandler.SEGMENTS_DESCRIPTOR_DIR_NAME));
    DruidStorageHandlerUtils.writeSegmentDescriptor(localFileSystem, dataSegment3, descriptorPath3);
    druidStorageHandler.commitInsertTable(tableMock, false);
    Assert.assertArrayEquals(Lists.newArrayList(DATA_SOURCE_NAME).toArray(), Lists.newArrayList(DruidStorageHandlerUtils.getAllDataSourceNames(connector, metadataStorageTablesConfig)).toArray());
    Assert.assertEquals(2, getUsedSegmentsList(connector, metadataStorageTablesConfig).size());
    // #8
    druidStorageHandler.commitInsertTable(tableMock, true);
    Assert.assertEquals(0, getUsedSegmentsList(connector, metadataStorageTablesConfig).size());
}

Example 7 with DataSegmentPusher

use of org.apache.druid.segment.loading.DataSegmentPusher in project hive by apache.

the class TestDruidStorageHandler method testInsertIntoAppendOneMorePartition.

@Test
public void testInsertIntoAppendOneMorePartition() throws MetaException, IOException {
    DerbyConnectorTestUtility connector = derbyConnectorRule.getConnector();
    MetadataStorageTablesConfig metadataStorageTablesConfig = derbyConnectorRule.metadataTablesConfigSupplier().get();
    druidStorageHandler.preCreateTable(tableMock);
    LocalFileSystem localFileSystem = FileSystem.getLocal(config);
    Path taskDirPath = new Path(tableWorkingPath, druidStorageHandler.makeStagingName());
    HdfsDataSegmentPusherConfig pusherConfig = new HdfsDataSegmentPusherConfig();
    pusherConfig.setStorageDirectory(config.get(String.valueOf(HiveConf.ConfVars.DRUID_SEGMENT_DIRECTORY)));
    DataSegmentPusher dataSegmentPusher = new HdfsDataSegmentPusher(pusherConfig, config, DruidStorageHandlerUtils.JSON_MAPPER);
    List<DataSegment> existingSegments = Collections.singletonList(createSegment(new Path(taskDirPath, DruidStorageHandlerUtils.INDEX_ZIP).toString(), new Interval(100, 150, DateTimeZone.UTC), "v0", new LinearShardSpec(0)));
    DruidStorageHandlerUtils.publishSegmentsAndCommit(connector, metadataStorageTablesConfig, DATA_SOURCE_NAME, existingSegments, true, config, dataSegmentPusher);
    DataSegment dataSegment = createSegment(new Path(taskDirPath, DruidStorageHandlerUtils.INDEX_ZIP).toString(), new Interval(100, 150, DateTimeZone.UTC), "v0", new LinearShardSpec(0));
    Path descriptorPath = DruidStorageHandlerUtils.makeSegmentDescriptorOutputPath(dataSegment, new Path(taskDirPath, DruidStorageHandler.SEGMENTS_DESCRIPTOR_DIR_NAME));
    DruidStorageHandlerUtils.writeSegmentDescriptor(localFileSystem, dataSegment, descriptorPath);
    druidStorageHandler.commitInsertTable(tableMock, false);
    Assert.assertArrayEquals(Lists.newArrayList(DATA_SOURCE_NAME).toArray(), Lists.newArrayList(DruidStorageHandlerUtils.getAllDataSourceNames(connector, metadataStorageTablesConfig)).toArray());
    final List<DataSegment> dataSegmentList = getUsedSegmentsList(connector, metadataStorageTablesConfig);
    Assert.assertEquals(2, dataSegmentList.size());
    DataSegment persistedSegment = dataSegmentList.get(1);
    Assert.assertEquals("v0", persistedSegment.getVersion());
    Assert.assertTrue(persistedSegment.getShardSpec() instanceof LinearShardSpec);
    Assert.assertEquals(1, persistedSegment.getShardSpec().getPartitionNum());
    Path expectedFinalHadoopPath = new Path(dataSegmentPusher.getPathForHadoop(), dataSegmentPusher.makeIndexPathName(persistedSegment, DruidStorageHandlerUtils.INDEX_ZIP));
    Assert.assertEquals(ImmutableMap.of("type", "hdfs", "path", expectedFinalHadoopPath.toString()), persistedSegment.getLoadSpec());
    Assert.assertEquals("dummySegmentData", FileUtils.readFileToString(new File(expectedFinalHadoopPath.toUri())));
}

Example 8 with DataSegmentPusher

use of org.apache.druid.segment.loading.DataSegmentPusher in project hive by apache.

the class TestDruidRecordWriter method testWrite.

// Test is failing due to Guava dependency, Druid 0.13.0 should have less dependency on Guava
@Ignore
@Test
public void testWrite() throws IOException, SegmentLoadingException {
    final String dataSourceName = "testDataSource";
    final File segmentOutputDir = temporaryFolder.newFolder();
    final File workingDir = temporaryFolder.newFolder();
    Configuration config = new Configuration();
    final InputRowParser inputRowParser = new MapInputRowParser(new TimeAndDimsParseSpec(new TimestampSpec(DruidConstants.DEFAULT_TIMESTAMP_COLUMN, "auto", null), new DimensionsSpec(ImmutableList.of(new StringDimensionSchema("host")), null, null)));
    final Map<String, Object> parserMap = objectMapper.convertValue(inputRowParser, new TypeReference<Map<String, Object>>() {
    });
    DataSchema dataSchema = new DataSchema(dataSourceName, parserMap, new AggregatorFactory[] { new LongSumAggregatorFactory("visited_sum", "visited_sum"), new HyperUniquesAggregatorFactory("unique_hosts", "unique_hosts") }, new UniformGranularitySpec(Granularities.DAY, Granularities.NONE, ImmutableList.of(INTERVAL_FULL)), null, objectMapper);
    IndexSpec indexSpec = new IndexSpec(new RoaringBitmapSerdeFactory(true), null, null, null);
    RealtimeTuningConfig tuningConfig = new RealtimeTuningConfig(null, null, null, null, temporaryFolder.newFolder(), null, null, null, null, indexSpec, null, null, 0, 0, null, null, 0L, null, null);
    LocalFileSystem localFileSystem = FileSystem.getLocal(config);
    DataSegmentPusher dataSegmentPusher = new LocalDataSegmentPusher(new LocalDataSegmentPusherConfig() {

        @Override
        public File getStorageDirectory() {
            return segmentOutputDir;
        }
    });
    Path segmentDescriptorPath = new Path(workingDir.getAbsolutePath(), DruidStorageHandler.SEGMENTS_DESCRIPTOR_DIR_NAME);
    DruidRecordWriter druidRecordWriter = new DruidRecordWriter(dataSchema, tuningConfig, dataSegmentPusher, 20, segmentDescriptorPath, localFileSystem);
    List<DruidWritable> druidWritables = expectedRows.stream().map(input -> new DruidWritable(ImmutableMap.<String, Object>builder().putAll(input).put(Constants.DRUID_TIMESTAMP_GRANULARITY_COL_NAME, Granularities.DAY.bucketStart(new DateTime((long) input.get(DruidConstants.DEFAULT_TIMESTAMP_COLUMN))).getMillis()).build())).collect(Collectors.toList());
    for (DruidWritable druidWritable : druidWritables) {
        druidRecordWriter.write(druidWritable);
    }
    druidRecordWriter.close(false);
    List<DataSegment> dataSegmentList = DruidStorageHandlerUtils.getCreatedSegments(segmentDescriptorPath, config);
    Assert.assertEquals(1, dataSegmentList.size());
    File tmpUnzippedSegmentDir = temporaryFolder.newFolder();
    new LocalDataSegmentPuller().getSegmentFiles(dataSegmentList.get(0), tmpUnzippedSegmentDir);
    final QueryableIndex queryableIndex = DruidStorageHandlerUtils.INDEX_IO.loadIndex(tmpUnzippedSegmentDir);
    QueryableIndexStorageAdapter adapter = new QueryableIndexStorageAdapter(queryableIndex);
    Firehose firehose = new IngestSegmentFirehose(ImmutableList.of(new WindowedStorageAdapter(adapter, adapter.getInterval())), null, ImmutableList.of("host"), ImmutableList.of("visited_sum", "unique_hosts"), null);
    List<InputRow> rows = Lists.newArrayList();
    while (firehose.hasMore()) {
        rows.add(firehose.nextRow());
    }
    verifyRows(expectedRows, rows);
}

Example 9 with DataSegmentPusher

use of org.apache.druid.segment.loading.DataSegmentPusher in project hive by apache.

the class TestDruidStorageHandler method testCommitInsertIntoTable.

@Test
public void testCommitInsertIntoTable() throws MetaException, IOException {
    DerbyConnectorTestUtility connector = derbyConnectorRule.getConnector();
    MetadataStorageTablesConfig metadataStorageTablesConfig = derbyConnectorRule.metadataTablesConfigSupplier().get();
    druidStorageHandler.preCreateTable(tableMock);
    LocalFileSystem localFileSystem = FileSystem.getLocal(config);
    Path taskDirPath = new Path(tableWorkingPath, druidStorageHandler.makeStagingName());
    List<DataSegment> existingSegments = Collections.singletonList(createSegment(new Path(taskDirPath, DruidStorageHandlerUtils.INDEX_ZIP).toString(), new Interval(100, 150, DateTimeZone.UTC), "v0", new LinearShardSpec(1)));
    HdfsDataSegmentPusherConfig pusherConfig = new HdfsDataSegmentPusherConfig();
    pusherConfig.setStorageDirectory(config.get(String.valueOf(HiveConf.ConfVars.DRUID_SEGMENT_DIRECTORY)));
    DataSegmentPusher dataSegmentPusher = new HdfsDataSegmentPusher(pusherConfig, config, DruidStorageHandlerUtils.JSON_MAPPER);
    DruidStorageHandlerUtils.publishSegmentsAndCommit(connector, metadataStorageTablesConfig, DATA_SOURCE_NAME, existingSegments, true, config, dataSegmentPusher);
    DataSegment dataSegment = createSegment(new Path(taskDirPath, DruidStorageHandlerUtils.INDEX_ZIP).toString(), new Interval(100, 150, DateTimeZone.UTC), "v1", new LinearShardSpec(0));
    Path descriptorPath = DruidStorageHandlerUtils.makeSegmentDescriptorOutputPath(dataSegment, new Path(taskDirPath, DruidStorageHandler.SEGMENTS_DESCRIPTOR_DIR_NAME));
    DruidStorageHandlerUtils.writeSegmentDescriptor(localFileSystem, dataSegment, descriptorPath);
    druidStorageHandler.commitInsertTable(tableMock, false);
    Assert.assertArrayEquals(Lists.newArrayList(DATA_SOURCE_NAME).toArray(), Lists.newArrayList(DruidStorageHandlerUtils.getAllDataSourceNames(connector, metadataStorageTablesConfig)).toArray());
    final List<DataSegment> dataSegmentList = getUsedSegmentsList(connector, metadataStorageTablesConfig);
    Assert.assertEquals(2, dataSegmentList.size());
    DataSegment persistedSegment = dataSegmentList.get(1);
    // Insert into appends to old version
    Assert.assertEquals("v0", persistedSegment.getVersion());
    Assert.assertTrue(persistedSegment.getShardSpec() instanceof LinearShardSpec);
    Assert.assertEquals(2, persistedSegment.getShardSpec().getPartitionNum());
    Path expectedFinalHadoopPath = new Path(dataSegmentPusher.getPathForHadoop(), dataSegmentPusher.makeIndexPathName(persistedSegment, DruidStorageHandlerUtils.INDEX_ZIP));
    Assert.assertEquals(ImmutableMap.of("type", "hdfs", "path", expectedFinalHadoopPath.toString()), persistedSegment.getLoadSpec());
    Assert.assertEquals("dummySegmentData", FileUtils.readFileToString(new File(expectedFinalHadoopPath.toUri())));
}

Example 10 with DataSegmentPusher

use of org.apache.druid.segment.loading.DataSegmentPusher in project hive by apache.

the class TestDruidStorageHandler method testCommitInsertIntoWithNonExtendableSegment.

@Test(expected = MetaException.class)
public void testCommitInsertIntoWithNonExtendableSegment() throws MetaException, IOException {
    DerbyConnectorTestUtility connector = derbyConnectorRule.getConnector();
    MetadataStorageTablesConfig metadataStorageTablesConfig = derbyConnectorRule.metadataTablesConfigSupplier().get();
    druidStorageHandler.preCreateTable(tableMock);
    LocalFileSystem localFileSystem = FileSystem.getLocal(config);
    Path taskDirPath = new Path(tableWorkingPath, druidStorageHandler.makeStagingName());
    List<DataSegment> existingSegments = Arrays.asList(createSegment(new Path(taskDirPath, "index_old_1.zip").toString(), new Interval(100, 150, DateTimeZone.UTC), "v0", NoneShardSpec.instance()), createSegment(new Path(taskDirPath, "index_old_2.zip").toString(), new Interval(200, 250, DateTimeZone.UTC), "v0", new LinearShardSpec(0)), createSegment(new Path(taskDirPath, "index_old_3.zip").toString(), new Interval(250, 300, DateTimeZone.UTC), "v0", new LinearShardSpec(0)));
    HdfsDataSegmentPusherConfig pusherConfig = new HdfsDataSegmentPusherConfig();
    pusherConfig.setStorageDirectory(taskDirPath.toString());
    DataSegmentPusher dataSegmentPusher = new HdfsDataSegmentPusher(pusherConfig, config, DruidStorageHandlerUtils.JSON_MAPPER);
    DruidStorageHandlerUtils.publishSegmentsAndCommit(connector, metadataStorageTablesConfig, DATA_SOURCE_NAME, existingSegments, true, config, dataSegmentPusher);
    // Try appending to non extendable shard spec
    DataSegment conflictingSegment = createSegment(new Path(taskDirPath, DruidStorageHandlerUtils.INDEX_ZIP).toString(), new Interval(100, 150, DateTimeZone.UTC), "v1", new LinearShardSpec(0));
    Path descriptorPath = DruidStorageHandlerUtils.makeSegmentDescriptorOutputPath(conflictingSegment, new Path(taskDirPath, DruidStorageHandler.SEGMENTS_DESCRIPTOR_DIR_NAME));
    DruidStorageHandlerUtils.writeSegmentDescriptor(localFileSystem, conflictingSegment, descriptorPath);
    druidStorageHandler.commitInsertTable(tableMock, false);
}