Search in sources :

Example 1 with CopyableFileWatermarkGenerator

use of org.apache.gobblin.data.management.copy.watermark.CopyableFileWatermarkGenerator in project incubator-gobblin by apache.

the class CopySource method getWorkunits.

/**
 * <ul>
 * Does the following:
 * <li>Instantiate a {@link DatasetsFinder}.
 * <li>Find all {@link Dataset} using {@link DatasetsFinder}.
 * <li>For each {@link CopyableDataset} get all {@link CopyEntity}s.
 * <li>Create a {@link WorkUnit} per {@link CopyEntity}.
 * </ul>
 *
 * <p>
 * In this implementation, one workunit is created for every {@link CopyEntity} found. But the extractor/converters
 * and writers are built to support multiple {@link CopyEntity}s per workunit
 * </p>
 *
 * @param state see {@link org.apache.gobblin.configuration.SourceState}
 * @return Work units for copying files.
 */
@Override
public List<WorkUnit> getWorkunits(final SourceState state) {
    this.metricContext = Instrumented.getMetricContext(state, CopySource.class);
    this.lineageInfo = LineageInfo.getLineageInfo(state.getBroker());
    try {
        DeprecationUtils.renameDeprecatedKeys(state, CopyConfiguration.MAX_COPY_PREFIX + "." + CopyResourcePool.ENTITIES_KEY, Lists.newArrayList(MAX_FILES_COPIED_KEY));
        final FileSystem sourceFs = HadoopUtils.getSourceFileSystem(state);
        final FileSystem targetFs = HadoopUtils.getWriterFileSystem(state, 1, 0);
        state.setProp(SlaEventKeys.SOURCE_URI, sourceFs.getUri());
        state.setProp(SlaEventKeys.DESTINATION_URI, targetFs.getUri());
        log.info("Identified source file system at {} and target file system at {}.", sourceFs.getUri(), targetFs.getUri());
        long maxSizePerBin = state.getPropAsLong(MAX_SIZE_MULTI_WORKUNITS, 0);
        long maxWorkUnitsPerMultiWorkUnit = state.getPropAsLong(MAX_WORK_UNITS_PER_BIN, 50);
        final long minWorkUnitWeight = Math.max(1, maxSizePerBin / maxWorkUnitsPerMultiWorkUnit);
        final Optional<CopyableFileWatermarkGenerator> watermarkGenerator = CopyableFileWatermarkHelper.getCopyableFileWatermarkGenerator(state);
        int maxThreads = state.getPropAsInt(MAX_CONCURRENT_LISTING_SERVICES, DEFAULT_MAX_CONCURRENT_LISTING_SERVICES);
        final CopyConfiguration copyConfiguration = CopyConfiguration.builder(targetFs, state.getProperties()).build();
        this.eventSubmitter = new EventSubmitter.Builder(this.metricContext, CopyConfiguration.COPY_PREFIX).build();
        DatasetsFinder<CopyableDatasetBase> datasetFinder = DatasetUtils.instantiateDatasetFinder(state.getProperties(), sourceFs, DEFAULT_DATASET_PROFILE_CLASS_KEY, this.eventSubmitter, state);
        IterableDatasetFinder<CopyableDatasetBase> iterableDatasetFinder = datasetFinder instanceof IterableDatasetFinder ? (IterableDatasetFinder<CopyableDatasetBase>) datasetFinder : new IterableDatasetFinderImpl<>(datasetFinder);
        Iterator<CopyableDatasetRequestor> requestorIteratorWithNulls = Iterators.transform(iterableDatasetFinder.getDatasetsIterator(), new CopyableDatasetRequestor.Factory(targetFs, copyConfiguration, log));
        Iterator<CopyableDatasetRequestor> requestorIterator = Iterators.filter(requestorIteratorWithNulls, Predicates.<CopyableDatasetRequestor>notNull());
        final SetMultimap<FileSet<CopyEntity>, WorkUnit> workUnitsMap = Multimaps.<FileSet<CopyEntity>, WorkUnit>synchronizedSetMultimap(HashMultimap.<FileSet<CopyEntity>, WorkUnit>create());
        RequestAllocator<FileSet<CopyEntity>> allocator = createRequestAllocator(copyConfiguration, maxThreads);
        Iterator<FileSet<CopyEntity>> prioritizedFileSets = allocator.allocateRequests(requestorIterator, copyConfiguration.getMaxToCopy());
        // Submit alertable events for unfulfilled requests
        submitUnfulfilledRequestEvents(allocator);
        Iterator<Callable<Void>> callableIterator = Iterators.transform(prioritizedFileSets, new Function<FileSet<CopyEntity>, Callable<Void>>() {

            @Nullable
            @Override
            public Callable<Void> apply(FileSet<CopyEntity> input) {
                return new FileSetWorkUnitGenerator((CopyableDatasetBase) input.getDataset(), input, state, workUnitsMap, watermarkGenerator, minWorkUnitWeight);
            }
        });
        try {
            List<Future<Void>> futures = new IteratorExecutor<>(callableIterator, maxThreads, ExecutorsUtils.newDaemonThreadFactory(Optional.of(log), Optional.of("Copy-file-listing-pool-%d"))).execute();
            for (Future<Void> future : futures) {
                try {
                    future.get();
                } catch (ExecutionException exc) {
                    log.error("Failed to get work units for dataset.", exc.getCause());
                }
            }
        } catch (InterruptedException ie) {
            log.error("Retrieval of work units was interrupted. Aborting.");
            return Lists.newArrayList();
        }
        log.info(String.format("Created %s workunits ", workUnitsMap.size()));
        copyConfiguration.getCopyContext().logCacheStatistics();
        if (state.contains(SIMULATE) && state.getPropAsBoolean(SIMULATE)) {
            log.info("Simulate mode enabled. Will not execute the copy.");
            for (Map.Entry<FileSet<CopyEntity>, Collection<WorkUnit>> entry : workUnitsMap.asMap().entrySet()) {
                log.info(String.format("Actions for dataset %s file set %s.", entry.getKey().getDataset().datasetURN(), entry.getKey().getName()));
                for (WorkUnit workUnit : entry.getValue()) {
                    CopyEntity copyEntity = deserializeCopyEntity(workUnit);
                    log.info(copyEntity.explain());
                }
            }
            return Lists.newArrayList();
        }
        List<? extends WorkUnit> workUnits = new WorstFitDecreasingBinPacking(maxSizePerBin).pack(Lists.newArrayList(workUnitsMap.values()), this.weighter);
        log.info(String.format("Bin packed work units. Initial work units: %d, packed work units: %d, max weight per bin: %d, " + "max work units per bin: %d.", workUnitsMap.size(), workUnits.size(), maxSizePerBin, maxWorkUnitsPerMultiWorkUnit));
        return ImmutableList.copyOf(workUnits);
    } catch (IOException e) {
        throw new RuntimeException(e);
    }
}
Also used : IterableDatasetFinder(org.apache.gobblin.dataset.IterableDatasetFinder) Callable(java.util.concurrent.Callable) WorstFitDecreasingBinPacking(org.apache.gobblin.util.binpacking.WorstFitDecreasingBinPacking) FileSystem(org.apache.hadoop.fs.FileSystem) ExecutionException(java.util.concurrent.ExecutionException) CopyableDatasetRequestor(org.apache.gobblin.data.management.partition.CopyableDatasetRequestor) FileSet(org.apache.gobblin.data.management.partition.FileSet) IOException(java.io.IOException) CopyableFileWatermarkGenerator(org.apache.gobblin.data.management.copy.watermark.CopyableFileWatermarkGenerator) Future(java.util.concurrent.Future) Collection(java.util.Collection) WorkUnit(org.apache.gobblin.source.workunit.WorkUnit) Map(java.util.Map) ImmutableMap(com.google.common.collect.ImmutableMap) Nullable(javax.annotation.Nullable)

Aggregations

ImmutableMap (com.google.common.collect.ImmutableMap)1 IOException (java.io.IOException)1 Collection (java.util.Collection)1 Map (java.util.Map)1 Callable (java.util.concurrent.Callable)1 ExecutionException (java.util.concurrent.ExecutionException)1 Future (java.util.concurrent.Future)1 Nullable (javax.annotation.Nullable)1 CopyableFileWatermarkGenerator (org.apache.gobblin.data.management.copy.watermark.CopyableFileWatermarkGenerator)1 CopyableDatasetRequestor (org.apache.gobblin.data.management.partition.CopyableDatasetRequestor)1 FileSet (org.apache.gobblin.data.management.partition.FileSet)1 IterableDatasetFinder (org.apache.gobblin.dataset.IterableDatasetFinder)1 WorkUnit (org.apache.gobblin.source.workunit.WorkUnit)1 WorstFitDecreasingBinPacking (org.apache.gobblin.util.binpacking.WorstFitDecreasingBinPacking)1 FileSystem (org.apache.hadoop.fs.FileSystem)1