Examples with TRIP_EXAMPLE_SCHEMA org.apache.hudi.common.testutils.HoodieTestDataGenerator.TRIP_EXAMPLE_SCHEMA used on opensource projects

Example 1 with TRIP_EXAMPLE_SCHEMA

use of org.apache.hudi.common.testutils.HoodieTestDataGenerator.TRIP_EXAMPLE_SCHEMA in project hudi by apache.

the class TestHoodieBackedMetadata method testReattemptOfFailedClusteringCommit.

/**
 * Lets say clustering commit succeeded in metadata table, but failed before committing to datatable.
 * Next time, when clustering kicks in, hudi will rollback pending clustering (in data table) and re-attempt the clustering with same
 * instant time. So, this test ensures the 2nd attempt succeeds with metadata enabled.
 * This is applicable to any table service where instant time is fixed. So, how many ever times the operation fails, re attempt will
 * be made with same commit time.
 * Tests uses clustering to test out the scenario.
 */
@Test
public void testReattemptOfFailedClusteringCommit() throws Exception {
    tableType = HoodieTableType.COPY_ON_WRITE;
    init(tableType);
    context = new HoodieSparkEngineContext(jsc);
    HoodieWriteConfig config = getSmallInsertWriteConfig(2000, TRIP_EXAMPLE_SCHEMA, 10, false);
    SparkRDDWriteClient client = getHoodieWriteClient(config);
    // Write 1 (Bulk insert)
    String newCommitTime = "0000001";
    List<HoodieRecord> records = dataGen.generateInserts(newCommitTime, 20);
    client.startCommitWithTime(newCommitTime);
    List<WriteStatus> writeStatuses = client.insert(jsc.parallelize(records, 1), newCommitTime).collect();
    assertNoWriteErrors(writeStatuses);
    validateMetadata(client);
    // Write 2 (inserts)
    newCommitTime = "0000002";
    client.startCommitWithTime(newCommitTime);
    records = dataGen.generateInserts(newCommitTime, 20);
    writeStatuses = client.insert(jsc.parallelize(records, 1), newCommitTime).collect();
    assertNoWriteErrors(writeStatuses);
    validateMetadata(client);
    // setup clustering config.
    HoodieClusteringConfig clusteringConfig = HoodieClusteringConfig.newBuilder().withClusteringMaxNumGroups(10).withClusteringSortColumns("_row_key").withInlineClustering(true).withClusteringTargetPartitions(0).withInlineClusteringNumCommits(1).build();
    HoodieWriteConfig newWriteConfig = getConfigBuilder(TRIP_EXAMPLE_SCHEMA, HoodieIndex.IndexType.BLOOM, HoodieFailedWritesCleaningPolicy.EAGER).withAutoCommit(false).withClusteringConfig(clusteringConfig).build();
    // trigger clustering
    SparkRDDWriteClient newClient = getHoodieWriteClient(newWriteConfig);
    String clusteringCommitTime = newClient.scheduleClustering(Option.empty()).get().toString();
    HoodieWriteMetadata<JavaRDD<WriteStatus>> clusterMetadata = newClient.cluster(clusteringCommitTime, true);
    // collect replaceFileIds for validation later.
    Set<HoodieFileGroupId> replacedFileIds = new HashSet<>();
    clusterMetadata.getPartitionToReplaceFileIds().entrySet().forEach(partitionFiles -> partitionFiles.getValue().stream().forEach(file -> replacedFileIds.add(new HoodieFileGroupId(partitionFiles.getKey(), file))));
    // trigger new write to mimic other writes succeeding before re-attempt.
    newCommitTime = "0000003";
    client.startCommitWithTime(newCommitTime);
    records = dataGen.generateInserts(newCommitTime, 20);
    writeStatuses = client.insert(jsc.parallelize(records, 1), newCommitTime).collect();
    assertNoWriteErrors(writeStatuses);
    validateMetadata(client);
    // manually remove clustering completed instant from .hoodie folder and to mimic succeeded clustering in metadata table, but failed in data table.
    FileCreateUtils.deleteReplaceCommit(basePath, clusteringCommitTime);
    HoodieWriteMetadata<JavaRDD<WriteStatus>> updatedClusterMetadata = newClient.cluster(clusteringCommitTime, true);
    metaClient.reloadActiveTimeline();
    Set<HoodieFileGroupId> updatedReplacedFileIds = new HashSet<>();
    updatedClusterMetadata.getPartitionToReplaceFileIds().entrySet().forEach(partitionFiles -> partitionFiles.getValue().stream().forEach(file -> updatedReplacedFileIds.add(new HoodieFileGroupId(partitionFiles.getKey(), file))));
    assertEquals(replacedFileIds, updatedReplacedFileIds);
    validateMetadata(client);
}

Example 2 with TRIP_EXAMPLE_SCHEMA

use of org.apache.hudi.common.testutils.HoodieTestDataGenerator.TRIP_EXAMPLE_SCHEMA in project hudi by apache.

the class TestHoodieClientOnCopyOnWriteStorage method testUpdateRejectForClustering.

@Test
public void testUpdateRejectForClustering() throws IOException {
    final String testPartitionPath = "2016/09/26";
    dataGen = new HoodieTestDataGenerator(new String[] { testPartitionPath });
    Properties props = new Properties();
    props.setProperty(ASYNC_CLUSTERING_ENABLE.key(), "true");
    HoodieWriteConfig config = getSmallInsertWriteConfig(100, TRIP_EXAMPLE_SCHEMA, dataGen.getEstimatedFileSizeInBytes(150), true, props);
    SparkRDDWriteClient client = getHoodieWriteClient(config);
    HoodieSparkCopyOnWriteTable table = (HoodieSparkCopyOnWriteTable) HoodieSparkTable.create(config, context, metaClient);
    // 1. insert to generate 2 file group
    String commitTime1 = "001";
    Pair<List<WriteStatus>, List<HoodieRecord>> upsertResult = insertBatchRecords(client, commitTime1, 600, 2);
    List<HoodieRecord> inserts1 = upsertResult.getValue();
    List<String> fileGroupIds1 = table.getFileSystemView().getAllFileGroups(testPartitionPath).map(fileGroup -> fileGroup.getFileGroupId().getFileId()).collect(Collectors.toList());
    assertEquals(2, fileGroupIds1.size());
    // 2. generate clustering plan for fileGroupIds1 file groups
    String commitTime2 = "002";
    List<List<FileSlice>> firstInsertFileSlicesList = table.getFileSystemView().getAllFileGroups(testPartitionPath).map(fileGroup -> fileGroup.getAllFileSlices().collect(Collectors.toList())).collect(Collectors.toList());
    List<FileSlice>[] fileSlices = (List<FileSlice>[]) firstInsertFileSlicesList.toArray(new List[firstInsertFileSlicesList.size()]);
    createRequestedReplaceInstant(this.metaClient, commitTime2, fileSlices);
    // 3. insert one record with no updating reject exception, and not merge the small file, just generate a new file group
    String commitTime3 = "003";
    insertBatchRecords(client, commitTime3, 1, 1).getKey();
    List<String> fileGroupIds2 = table.getFileSystemView().getAllFileGroups(testPartitionPath).map(fileGroup -> fileGroup.getFileGroupId().getFileId()).collect(Collectors.toList());
    assertEquals(3, fileGroupIds2.size());
    // 4. update one record for the clustering two file groups, throw reject update exception
    String commitTime4 = "004";
    client.startCommitWithTime(commitTime4);
    List<HoodieRecord> insertsAndUpdates3 = new ArrayList<>();
    insertsAndUpdates3.addAll(dataGen.generateUpdates(commitTime4, inserts1));
    String assertMsg = String.format("Not allowed to update the clustering files in partition: %s " + "For pending clustering operations, we are not going to support update for now.", testPartitionPath);
    assertThrows(HoodieUpsertException.class, () -> {
        writeClient.upsert(jsc.parallelize(insertsAndUpdates3, 1), commitTime3).collect();
    }, assertMsg);
    // 5. insert one record with no updating reject exception, will merge the small file
    String commitTime5 = "005";
    List<WriteStatus> statuses = insertBatchRecords(client, commitTime5, 1, 1).getKey();
    fileGroupIds2.removeAll(fileGroupIds1);
    assertEquals(fileGroupIds2.get(0), statuses.get(0).getFileId());
    List<String> firstInsertFileGroupIds4 = table.getFileSystemView().getAllFileGroups(testPartitionPath).map(fileGroup -> fileGroup.getFileGroupId().getFileId()).collect(Collectors.toList());
    assertEquals(3, firstInsertFileGroupIds4.size());
}