Search in sources :

Example 6 with TableFileSystemView

use of org.apache.hudi.common.table.view.TableFileSystemView in project hudi by apache.

the class TestCleaner method testInsertAndCleanByVersions.

/**
 * Test Helper for Cleaning by versions logic from HoodieWriteClient API perspective.
 *
 * @param insertFn Insert API to be tested
 * @param upsertFn Upsert API to be tested
 * @param isPreppedAPI Flag to indicate if a prepped-version is used. If true, a wrapper function will be used during
 *        record generation to also tag the regards (de-dupe is implicit as we use unique record-gen APIs)
 * @throws Exception in case of errors
 */
private void testInsertAndCleanByVersions(Function3<JavaRDD<WriteStatus>, SparkRDDWriteClient, JavaRDD<HoodieRecord>, String> insertFn, Function3<JavaRDD<WriteStatus>, SparkRDDWriteClient, JavaRDD<HoodieRecord>, String> upsertFn, boolean isPreppedAPI) throws Exception {
    // keep upto 2 versions for each file
    int maxVersions = 2;
    HoodieWriteConfig cfg = getConfigBuilder().withCompactionConfig(HoodieCompactionConfig.newBuilder().withCleanerPolicy(HoodieCleaningPolicy.KEEP_LATEST_FILE_VERSIONS).retainFileVersions(maxVersions).build()).withParallelism(1, 1).withBulkInsertParallelism(1).withFinalizeWriteParallelism(1).withDeleteParallelism(1).withConsistencyGuardConfig(ConsistencyGuardConfig.newBuilder().withConsistencyCheckEnabled(true).build()).build();
    try (SparkRDDWriteClient client = getHoodieWriteClient(cfg)) {
        final Function2<List<HoodieRecord>, String, Integer> recordInsertGenWrappedFunction = generateWrapRecordsFn(isPreppedAPI, cfg, dataGen::generateInserts);
        final Function2<List<HoodieRecord>, String, Integer> recordUpsertGenWrappedFunction = generateWrapRecordsFn(isPreppedAPI, cfg, dataGen::generateUniqueUpdates);
        insertFirstBigBatchForClientCleanerTest(cfg, client, recordInsertGenWrappedFunction, insertFn, HoodieCleaningPolicy.KEEP_LATEST_FILE_VERSIONS);
        Map<HoodieFileGroupId, FileSlice> compactionFileIdToLatestFileSlice = new HashMap<>();
        metaClient = HoodieTableMetaClient.reload(metaClient);
        HoodieTable table = HoodieSparkTable.create(getConfig(), context, metaClient);
        for (String partitionPath : dataGen.getPartitionPaths()) {
            TableFileSystemView fsView = table.getFileSystemView();
            Option<Boolean> added = Option.fromJavaOptional(fsView.getAllFileGroups(partitionPath).findFirst().map(fg -> {
                fg.getLatestFileSlice().map(fs -> compactionFileIdToLatestFileSlice.put(fg.getFileGroupId(), fs));
                return true;
            }));
            if (added.isPresent()) {
                // Select only one file-group for compaction
                break;
            }
        }
        // Create workload with selected file-slices
        List<Pair<String, FileSlice>> partitionFileSlicePairs = compactionFileIdToLatestFileSlice.entrySet().stream().map(e -> Pair.of(e.getKey().getPartitionPath(), e.getValue())).collect(Collectors.toList());
        HoodieCompactionPlan compactionPlan = CompactionUtils.buildFromFileSlices(partitionFileSlicePairs, Option.empty(), Option.empty());
        List<String> instantTimes = makeIncrementalCommitTimes(9, 1, 10);
        String compactionTime = instantTimes.get(0);
        table.getActiveTimeline().saveToCompactionRequested(new HoodieInstant(State.REQUESTED, HoodieTimeline.COMPACTION_ACTION, compactionTime), TimelineMetadataUtils.serializeCompactionPlan(compactionPlan));
        instantTimes = instantTimes.subList(1, instantTimes.size());
        // remaining.
        for (String newInstantTime : instantTimes) {
            try {
                client.startCommitWithTime(newInstantTime);
                List<HoodieRecord> records = recordUpsertGenWrappedFunction.apply(newInstantTime, 100);
                List<WriteStatus> statuses = upsertFn.apply(client, jsc.parallelize(records, 1), newInstantTime).collect();
                // Verify there are no errors
                assertNoWriteErrors(statuses);
                metaClient = HoodieTableMetaClient.reload(metaClient);
                table = HoodieSparkTable.create(getConfig(), context, metaClient);
                HoodieTimeline timeline = table.getMetaClient().getCommitsTimeline();
                TableFileSystemView fsView = table.getFileSystemView();
                // Need to ensure the following
                for (String partitionPath : dataGen.getPartitionPaths()) {
                    // compute all the versions of all files, from time 0
                    HashMap<String, TreeSet<String>> fileIdToVersions = new HashMap<>();
                    for (HoodieInstant entry : timeline.getInstants().collect(Collectors.toList())) {
                        HoodieCommitMetadata commitMetadata = HoodieCommitMetadata.fromBytes(timeline.getInstantDetails(entry).get(), HoodieCommitMetadata.class);
                        for (HoodieWriteStat wstat : commitMetadata.getWriteStats(partitionPath)) {
                            if (!fileIdToVersions.containsKey(wstat.getFileId())) {
                                fileIdToVersions.put(wstat.getFileId(), new TreeSet<>());
                            }
                            fileIdToVersions.get(wstat.getFileId()).add(FSUtils.getCommitTime(new Path(wstat.getPath()).getName()));
                        }
                    }
                    List<HoodieFileGroup> fileGroups = fsView.getAllFileGroups(partitionPath).collect(Collectors.toList());
                    for (HoodieFileGroup fileGroup : fileGroups) {
                        if (compactionFileIdToLatestFileSlice.containsKey(fileGroup.getFileGroupId())) {
                            // Ensure latest file-slice selected for compaction is retained
                            Option<HoodieBaseFile> dataFileForCompactionPresent = Option.fromJavaOptional(fileGroup.getAllBaseFiles().filter(df -> {
                                return compactionFileIdToLatestFileSlice.get(fileGroup.getFileGroupId()).getBaseInstantTime().equals(df.getCommitTime());
                            }).findAny());
                            assertTrue(dataFileForCompactionPresent.isPresent(), "Data File selected for compaction is retained");
                        } else {
                            // file has no more than max versions
                            String fileId = fileGroup.getFileGroupId().getFileId();
                            List<HoodieBaseFile> dataFiles = fileGroup.getAllBaseFiles().collect(Collectors.toList());
                            assertTrue(dataFiles.size() <= maxVersions, "fileId " + fileId + " has more than " + maxVersions + " versions");
                            // Each file, has the latest N versions (i.e cleaning gets rid of older versions)
                            List<String> commitedVersions = new ArrayList<>(fileIdToVersions.get(fileId));
                            for (int i = 0; i < dataFiles.size(); i++) {
                                assertEquals((dataFiles.get(i)).getCommitTime(), commitedVersions.get(commitedVersions.size() - 1 - i), "File " + fileId + " does not have latest versions on commits" + commitedVersions);
                            }
                        }
                    }
                }
            } catch (IOException ioe) {
                throw new RuntimeException(ioe);
            }
        }
    }
}
Also used : Arrays(java.util.Arrays) HoodieFileGroup(org.apache.hudi.common.model.HoodieFileGroup) CleanPlanner(org.apache.hudi.table.action.clean.CleanPlanner) HoodieFileStatus(org.apache.hudi.avro.model.HoodieFileStatus) Map(java.util.Map) HoodieRollbackMetadata(org.apache.hudi.avro.model.HoodieRollbackMetadata) HoodieFileGroupId(org.apache.hudi.common.model.HoodieFileGroupId) HoodieActiveTimeline(org.apache.hudi.common.table.timeline.HoodieActiveTimeline) Awaitility.await(org.awaitility.Awaitility.await) HoodieCleanerPlan(org.apache.hudi.avro.model.HoodieCleanerPlan) HoodieClusteringPlan(org.apache.hudi.avro.model.HoodieClusteringPlan) Set(java.util.Set) Arguments(org.junit.jupiter.params.provider.Arguments) HoodieIndex(org.apache.hudi.index.HoodieIndex) StandardCharsets(java.nio.charset.StandardCharsets) HoodieBaseFile(org.apache.hudi.common.model.HoodieBaseFile) Stream(java.util.stream.Stream) FileSystemViewStorageConfig(org.apache.hudi.common.table.view.FileSystemViewStorageConfig) Assertions.assertTrue(org.junit.jupiter.api.Assertions.assertTrue) TableFileSystemView(org.apache.hudi.common.table.view.TableFileSystemView) HoodieClientTestBase(org.apache.hudi.testutils.HoodieClientTestBase) Assertions.assertNotNull(org.junit.jupiter.api.Assertions.assertNotNull) HoodieCleaningPolicy(org.apache.hudi.common.model.HoodieCleaningPolicy) CleanPlanMigrator(org.apache.hudi.common.table.timeline.versioning.clean.CleanPlanMigrator) Assertions.assertNull(org.junit.jupiter.api.Assertions.assertNull) Option(org.apache.hudi.common.util.Option) CleanPlanV1MigrationHandler(org.apache.hudi.common.table.timeline.versioning.clean.CleanPlanV1MigrationHandler) TreeSet(java.util.TreeSet) ArrayList(java.util.ArrayList) FSDataOutputStream(org.apache.hadoop.fs.FSDataOutputStream) StringUtils(org.apache.hudi.common.util.StringUtils) CleanerUtils(org.apache.hudi.common.util.CleanerUtils) HoodieTestCommitGenerator.getBaseFilename(org.apache.hudi.HoodieTestCommitGenerator.getBaseFilename) Assertions.assertEquals(org.junit.jupiter.api.Assertions.assertEquals) JavaRDD(org.apache.spark.api.java.JavaRDD) HoodieCleanStat(org.apache.hudi.common.HoodieCleanStat) HoodieMetadataConfig(org.apache.hudi.common.config.HoodieMetadataConfig) ValueSource(org.junit.jupiter.params.provider.ValueSource) ConsistencyGuardConfig(org.apache.hudi.common.fs.ConsistencyGuardConfig) HoodieRecord(org.apache.hudi.common.model.HoodieRecord) Assertions.assertNoWriteErrors(org.apache.hudi.testutils.Assertions.assertNoWriteErrors) HoodieWriteConfig(org.apache.hudi.config.HoodieWriteConfig) Files(java.nio.file.Files) HoodieTestTable(org.apache.hudi.common.testutils.HoodieTestTable) HoodieCommitMetadata(org.apache.hudi.common.model.HoodieCommitMetadata) IOException(java.io.IOException) File(java.io.File) HoodieCompactionConfig(org.apache.hudi.config.HoodieCompactionConfig) HoodieClusteringGroup(org.apache.hudi.avro.model.HoodieClusteringGroup) ParameterizedTest(org.junit.jupiter.params.ParameterizedTest) HoodieCleanMetadata(org.apache.hudi.avro.model.HoodieCleanMetadata) Paths(java.nio.file.Paths) HoodieTableMetadataWriter(org.apache.hudi.metadata.HoodieTableMetadataWriter) HoodieIOException(org.apache.hudi.exception.HoodieIOException) HoodieCleanPartitionMetadata(org.apache.hudi.avro.model.HoodieCleanPartitionMetadata) HoodieTestUtils(org.apache.hudi.common.testutils.HoodieTestUtils) HoodieFailedWritesCleaningPolicy(org.apache.hudi.common.model.HoodieFailedWritesCleaningPolicy) HoodieInstant(org.apache.hudi.common.table.timeline.HoodieInstant) CollectionUtils(org.apache.hudi.common.util.CollectionUtils) HoodieTestTable.makeIncrementalCommitTimes(org.apache.hudi.common.testutils.HoodieTestTable.makeIncrementalCommitTimes) HoodieMetadataTestTable(org.apache.hudi.common.testutils.HoodieMetadataTestTable) Logger(org.apache.log4j.Logger) HoodieTableType(org.apache.hudi.common.model.HoodieTableType) Assertions.assertFalse(org.junit.jupiter.api.Assertions.assertFalse) SparkHoodieBackedTableMetadataWriter(org.apache.hudi.metadata.SparkHoodieBackedTableMetadataWriter) Path(org.apache.hadoop.fs.Path) MethodSource(org.junit.jupiter.params.provider.MethodSource) IOType(org.apache.hudi.common.model.IOType) Predicate(java.util.function.Predicate) TimelineMetadataUtils(org.apache.hudi.common.table.timeline.TimelineMetadataUtils) UUID(java.util.UUID) Collectors(java.util.stream.Collectors) Tuple3(scala.Tuple3) HoodieClusteringStrategy(org.apache.hudi.avro.model.HoodieClusteringStrategy) Test(org.junit.jupiter.api.Test) List(java.util.List) HoodieWriteStat(org.apache.hudi.common.model.HoodieWriteStat) CleanMetadataMigrator(org.apache.hudi.common.table.timeline.versioning.clean.CleanMetadataMigrator) HoodieCompactionPlan(org.apache.hudi.avro.model.HoodieCompactionPlan) WriteOperationType(org.apache.hudi.common.model.WriteOperationType) DEFAULT_PARTITION_PATHS(org.apache.hudi.common.testutils.HoodieTestUtils.DEFAULT_PARTITION_PATHS) CompactionUtils(org.apache.hudi.common.util.CompactionUtils) FileSlice(org.apache.hudi.common.model.FileSlice) HashMap(java.util.HashMap) State(org.apache.hudi.common.table.timeline.HoodieInstant.State) HashSet(java.util.HashSet) HoodieRequestedReplaceMetadata(org.apache.hudi.avro.model.HoodieRequestedReplaceMetadata) HoodieTableMetaClient(org.apache.hudi.common.table.HoodieTableMetaClient) HoodieTimeline(org.apache.hudi.common.table.timeline.HoodieTimeline) BootstrapFileMapping(org.apache.hudi.common.model.BootstrapFileMapping) TestBootstrapIndex(org.apache.hudi.common.bootstrap.TestBootstrapIndex) HoodieActionInstant(org.apache.hudi.avro.model.HoodieActionInstant) HoodieReplaceCommitMetadata(org.apache.hudi.common.model.HoodieReplaceCommitMetadata) TimeUnit(java.util.concurrent.TimeUnit) WriteStatus(org.apache.hudi.client.WriteStatus) HoodieTestTable.makeNewCommitTime(org.apache.hudi.common.testutils.HoodieTestTable.makeNewCommitTime) SparkRDDWriteClient(org.apache.hudi.client.SparkRDDWriteClient) SparkHoodieIndexFactory(org.apache.hudi.index.SparkHoodieIndexFactory) HoodieSliceInfo(org.apache.hudi.avro.model.HoodieSliceInfo) LogManager(org.apache.log4j.LogManager) Collections(java.util.Collections) FSUtils(org.apache.hudi.common.fs.FSUtils) Pair(org.apache.hudi.common.util.collection.Pair) HoodieBaseFile(org.apache.hudi.common.model.HoodieBaseFile) HashMap(java.util.HashMap) FileSlice(org.apache.hudi.common.model.FileSlice) HoodieRecord(org.apache.hudi.common.model.HoodieRecord) HoodieTimeline(org.apache.hudi.common.table.timeline.HoodieTimeline) ArrayList(java.util.ArrayList) HoodieCommitMetadata(org.apache.hudi.common.model.HoodieCommitMetadata) HoodieCompactionPlan(org.apache.hudi.avro.model.HoodieCompactionPlan) HoodieFileGroupId(org.apache.hudi.common.model.HoodieFileGroupId) TreeSet(java.util.TreeSet) ArrayList(java.util.ArrayList) List(java.util.List) TableFileSystemView(org.apache.hudi.common.table.view.TableFileSystemView) WriteStatus(org.apache.hudi.client.WriteStatus) Pair(org.apache.hudi.common.util.collection.Pair) HoodieInstant(org.apache.hudi.common.table.timeline.HoodieInstant) Path(org.apache.hadoop.fs.Path) SparkRDDWriteClient(org.apache.hudi.client.SparkRDDWriteClient) HoodieWriteStat(org.apache.hudi.common.model.HoodieWriteStat) HoodieWriteConfig(org.apache.hudi.config.HoodieWriteConfig) IOException(java.io.IOException) HoodieIOException(org.apache.hudi.exception.HoodieIOException) HoodieFileGroup(org.apache.hudi.common.model.HoodieFileGroup)

Aggregations

IOException (java.io.IOException)6 ArrayList (java.util.ArrayList)6 List (java.util.List)6 HoodieRecord (org.apache.hudi.common.model.HoodieRecord)6 Arrays (java.util.Arrays)5 Collections (java.util.Collections)5 Map (java.util.Map)5 Collectors (java.util.stream.Collectors)5 Path (org.apache.hadoop.fs.Path)5 SparkRDDWriteClient (org.apache.hudi.client.SparkRDDWriteClient)5 FileSlice (org.apache.hudi.common.model.FileSlice)5 HoodieFileGroup (org.apache.hudi.common.model.HoodieFileGroup)5 HoodieTableType (org.apache.hudi.common.model.HoodieTableType)5 HoodieTableMetaClient (org.apache.hudi.common.table.HoodieTableMetaClient)5 HoodieInstant (org.apache.hudi.common.table.timeline.HoodieInstant)5 HashMap (java.util.HashMap)4 FileStatus (org.apache.hadoop.fs.FileStatus)4 TableFileSystemView (org.apache.hudi.common.table.view.TableFileSystemView)4 HoodieWriteConfig (org.apache.hudi.config.HoodieWriteConfig)4 HashSet (java.util.HashSet)3