Examples with SampleDataForSplitPointsJobFactory uk.gov.gchq.gaffer.accumulostore.operation.hdfs.handler.job.factory.SampleDataForSplitPointsJobFactory used on opensource projects

Example 1 with SampleDataForSplitPointsJobFactory

use of uk.gov.gchq.gaffer.accumulostore.operation.hdfs.handler.job.factory.SampleDataForSplitPointsJobFactory in project Gaffer by gchq.

the class SampleDataAndCreateSplitsFileTool method run.

@Override
public int run(final String[] strings) throws OperationException {
    try {
        LOGGER.info("Creating job using SampleDataForSplitPointsJobFactory");
        job = new SampleDataForSplitPointsJobFactory().createJob(operation, store);
    } catch (final IOException e) {
        LOGGER.error("Failed to create Hadoop job: {}", e.getMessage());
        throw new OperationException("Failed to create the Hadoop job: " + e.getMessage(), e);
    }
    try {
        LOGGER.info("Running SampleDataForSplitPoints job (job name is {})", job.getJobName());
        job.waitForCompletion(true);
    } catch (final IOException | InterruptedException | ClassNotFoundException e) {
        LOGGER.error("Exception running job: {}", e.getMessage());
        throw new OperationException("Error while waiting for job to complete: " + e.getMessage(), e);
    }
    try {
        if (!job.isSuccessful()) {
            LOGGER.error("Job was not successful (job name is {})", job.getJobName());
            throw new OperationException("Error running job");
        }
    } catch (final IOException e) {
        LOGGER.error("Exception running job: {}", e.getMessage());
        throw new OperationException("Error running job" + e.getMessage(), e);
    }
    // Find the number of records output
    // NB In the following line use mapred.Task.Counter.REDUCE_OUTPUT_RECORDS rather than
    // mapreduce.TaskCounter.REDUCE_OUTPUT_RECORDS as this is more compatible with earlier
    // versions of Hadoop.
    Counter counter;
    try {
        counter = job.getCounters().findCounter(Task.Counter.REDUCE_OUTPUT_RECORDS);
        LOGGER.info("Number of records output = {}", counter);
    } catch (final IOException e) {
        LOGGER.error("Failed to get counter org.apache.hadoop.mapred.Task.Counter.REDUCE_OUTPUT_RECORDS from job: {}", e.getMessage());
        throw new OperationException("Failed to get counter: " + Task.Counter.REDUCE_OUTPUT_RECORDS, e);
    }
    int numberTabletServers;
    try {
        numberTabletServers = store.getConnection().instanceOperations().getTabletServers().size();
        LOGGER.info("Number of tablet servers is {}", numberTabletServers);
    } catch (final StoreException e) {
        LOGGER.error("Exception thrown getting number of tablet servers: {}", e.getMessage());
        throw new OperationException(e.getMessage(), e);
    }
    long outputEveryNthRecord = counter.getValue() / (numberTabletServers - 1);
    final Path resultsFile = new Path(operation.getOutputPath(), "part-r-00000");
    LOGGER.info("Will output every {}-th record from {}", outputEveryNthRecord, resultsFile);
    // Read through resulting file, pick out the split points and write to file.
    final Configuration conf = getConf();
    final FileSystem fs;
    try {
        fs = FileSystem.get(conf);
    } catch (final IOException e) {
        LOGGER.error("Exception getting filesystem: {}", e.getMessage());
        throw new OperationException("Failed to get filesystem from configuration: " + e.getMessage(), e);
    }
    LOGGER.info("Writing splits to {}", operation.getResultingSplitsFilePath());
    final Key key = new Key();
    final Value value = new Value();
    long count = 0;
    int numberSplitPointsOutput = 0;
    try (final SequenceFile.Reader reader = new SequenceFile.Reader(fs, resultsFile, conf);
        final PrintStream splitsWriter = new PrintStream(new BufferedOutputStream(fs.create(new Path(operation.getResultingSplitsFilePath()), true)), false, CommonConstants.UTF_8)) {
        while (reader.next(key, value) && numberSplitPointsOutput < numberTabletServers - 1) {
            count++;
            if (count % outputEveryNthRecord == 0) {
                LOGGER.debug("Outputting split point number {} ({})", numberSplitPointsOutput, Base64.encodeBase64(key.getRow().getBytes()));
                numberSplitPointsOutput++;
                splitsWriter.println(new String(Base64.encodeBase64(key.getRow().getBytes()), CommonConstants.UTF_8));
            }
        }
        LOGGER.info("Total number of records read was {}", count);
    } catch (final IOException e) {
        LOGGER.error("Exception reading results file and outputting split points: {}", e.getMessage());
        throw new OperationException(e.getMessage(), e);
    }
    try {
        fs.delete(resultsFile, true);
        LOGGER.info("Deleted the results file {}", resultsFile);
    } catch (final IOException e) {
        LOGGER.error("Failed to delete the results file {}", resultsFile);
        throw new OperationException("Failed to delete the results file: " + e.getMessage(), e);
    }
    return SUCCESS_RESPONSE;
}