Search in sources :

Example 6 with HadoopTaskCancelledException

use of org.apache.ignite.internal.processors.hadoop.HadoopTaskCancelledException in project ignite by apache.

the class HadoopV1MapTask method run.

/** {@inheritDoc} */
@SuppressWarnings("unchecked")
@Override
public void run(HadoopTaskContext taskCtx) throws IgniteCheckedException {
    HadoopJobEx job = taskCtx.job();
    HadoopV2TaskContext taskCtx0 = (HadoopV2TaskContext) taskCtx;
    if (taskCtx.taskInfo().hasMapperIndex())
        HadoopMapperUtils.mapperIndex(taskCtx.taskInfo().mapperIndex());
    else
        HadoopMapperUtils.clearMapperIndex();
    try {
        JobConf jobConf = taskCtx0.jobConf();
        InputFormat inFormat = jobConf.getInputFormat();
        HadoopInputSplit split = info().inputSplit();
        InputSplit nativeSplit;
        if (split instanceof HadoopFileBlock) {
            HadoopFileBlock block = (HadoopFileBlock) split;
            nativeSplit = new FileSplit(new Path(block.file().toString()), block.start(), block.length(), EMPTY_HOSTS);
        } else
            nativeSplit = (InputSplit) taskCtx0.getNativeSplit(split);
        assert nativeSplit != null;
        Reporter reporter = new HadoopV1Reporter(taskCtx);
        HadoopV1OutputCollector collector = null;
        try {
            collector = collector(jobConf, taskCtx0, !job.info().hasCombiner() && !job.info().hasReducer(), fileName(), taskCtx0.attemptId());
            RecordReader reader = inFormat.getRecordReader(nativeSplit, jobConf, reporter);
            Mapper mapper = ReflectionUtils.newInstance(jobConf.getMapperClass(), jobConf);
            Object key = reader.createKey();
            Object val = reader.createValue();
            assert mapper != null;
            try {
                try {
                    while (reader.next(key, val)) {
                        if (isCancelled())
                            throw new HadoopTaskCancelledException("Map task cancelled.");
                        mapper.map(key, val, collector, reporter);
                    }
                    taskCtx.onMapperFinished();
                } finally {
                    mapper.close();
                }
            } finally {
                collector.closeWriter();
            }
            collector.commit();
        } catch (Exception e) {
            if (collector != null)
                collector.abort();
            throw new IgniteCheckedException(e);
        }
    } finally {
        HadoopMapperUtils.clearMapperIndex();
    }
}
Also used : Path(org.apache.hadoop.fs.Path) Reporter(org.apache.hadoop.mapred.Reporter) RecordReader(org.apache.hadoop.mapred.RecordReader) HadoopInputSplit(org.apache.ignite.hadoop.HadoopInputSplit) HadoopFileBlock(org.apache.ignite.internal.processors.hadoop.HadoopFileBlock) FileSplit(org.apache.hadoop.mapred.FileSplit) IgniteCheckedException(org.apache.ignite.IgniteCheckedException) HadoopTaskCancelledException(org.apache.ignite.internal.processors.hadoop.HadoopTaskCancelledException) Mapper(org.apache.hadoop.mapred.Mapper) IgniteCheckedException(org.apache.ignite.IgniteCheckedException) HadoopJobEx(org.apache.ignite.internal.processors.hadoop.HadoopJobEx) InputFormat(org.apache.hadoop.mapred.InputFormat) HadoopTaskCancelledException(org.apache.ignite.internal.processors.hadoop.HadoopTaskCancelledException) JobConf(org.apache.hadoop.mapred.JobConf) HadoopInputSplit(org.apache.ignite.hadoop.HadoopInputSplit) InputSplit(org.apache.hadoop.mapred.InputSplit) HadoopV2TaskContext(org.apache.ignite.internal.processors.hadoop.impl.v2.HadoopV2TaskContext)

Example 7 with HadoopTaskCancelledException

use of org.apache.ignite.internal.processors.hadoop.HadoopTaskCancelledException in project ignite by apache.

the class HadoopV1Task method collector.

/**
     *
     * @param jobConf Job configuration.
     * @param taskCtx Task context.
     * @param directWrite Direct write flag.
     * @param fileName File name.
     * @param attempt Attempt of task.
     * @return Collector.
     * @throws IOException In case of IO exception.
     */
protected HadoopV1OutputCollector collector(JobConf jobConf, HadoopV2TaskContext taskCtx, boolean directWrite, @Nullable String fileName, TaskAttemptID attempt) throws IOException {
    HadoopV1OutputCollector collector = new HadoopV1OutputCollector(jobConf, taskCtx, directWrite, fileName, attempt) {

        /** {@inheritDoc} */
        @Override
        public void collect(Object key, Object val) throws IOException {
            if (cancelled)
                throw new HadoopTaskCancelledException("Task cancelled.");
            super.collect(key, val);
        }
    };
    collector.setup();
    return collector;
}
Also used : HadoopTaskCancelledException(org.apache.ignite.internal.processors.hadoop.HadoopTaskCancelledException)

Aggregations

HadoopTaskCancelledException (org.apache.ignite.internal.processors.hadoop.HadoopTaskCancelledException)7 IgniteCheckedException (org.apache.ignite.IgniteCheckedException)3 JobConf (org.apache.hadoop.mapred.JobConf)2 HadoopJobEx (org.apache.ignite.internal.processors.hadoop.HadoopJobEx)2 HadoopTaskInput (org.apache.ignite.internal.processors.hadoop.HadoopTaskInput)2 HadoopV2TaskContext (org.apache.ignite.internal.processors.hadoop.impl.v2.HadoopV2TaskContext)2 Path (org.apache.hadoop.fs.Path)1 FileSplit (org.apache.hadoop.mapred.FileSplit)1 InputFormat (org.apache.hadoop.mapred.InputFormat)1 InputSplit (org.apache.hadoop.mapred.InputSplit)1 Mapper (org.apache.hadoop.mapred.Mapper)1 RecordReader (org.apache.hadoop.mapred.RecordReader)1 Reducer (org.apache.hadoop.mapred.Reducer)1 Reporter (org.apache.hadoop.mapred.Reporter)1 HadoopInputSplit (org.apache.ignite.hadoop.HadoopInputSplit)1 HadoopFileBlock (org.apache.ignite.internal.processors.hadoop.HadoopFileBlock)1 HadoopTaskInfo (org.apache.ignite.internal.processors.hadoop.HadoopTaskInfo)1 HadoopTaskOutput (org.apache.ignite.internal.processors.hadoop.HadoopTaskOutput)1 HadoopPerformanceCounter (org.apache.ignite.internal.processors.hadoop.counter.HadoopPerformanceCounter)1