def run(self, job_name, bucket_name, filenames):
sort_mappers = []
for i in range(len(filenames)):
filenames_only = util.strip_prefix_from_items("/%s/" % bucket_name,
filenames[i])
sort_mapper = yield mapper_pipeline.MapperPipeline(
"%s-shuffle-sort-%s" % (job_name, str(i)),
__name__ + "._sort_records_map",
__name__ + "._BatchGCSRecordsReader",
None,
{
"input_reader": {
"bucket_name": bucket_name,
"objects": filenames_only,
},
},
shards=1)
sort_mappers.append(sort_mapper)
with pipeline.After(*sort_mappers):
job_ids = yield pipeline_common.Append(*[mapper.job_id for mapper in
sort_mappers])
result = yield _CollectOutputFiles(job_ids)
with pipeline.After(result):
yield _CleanupOutputFiles(job_ids)
yield pipeline_common.Return(result)
def run(self,
job_name,
mapper_spec,
reducer_spec,
input_reader_spec,
output_writer_spec=None,
mapper_params=None,
reducer_params=None,
shards=None,
combiner_spec=None):
map_pipeline = yield MapPipeline(job_name,
mapper_spec,
input_reader_spec,
params=mapper_params,
shards=shards)
shuffler_pipeline = yield ShufflePipeline(job_name, map_pipeline)
reducer_pipeline = yield ReducePipeline(job_name,
reducer_spec,
output_writer_spec,
reducer_params,
shuffler_pipeline,
combiner_spec=combiner_spec)
with pipeline.After(reducer_pipeline):
all_temp_files = yield pipeline_common.Extend(
map_pipeline, shuffler_pipeline)
yield CleanupPipeline(all_temp_files)
yield pipeline_common.Return(reducer_pipeline)
def run(self, job_name, filenames):
hashed_files = yield _HashPipeline(job_name, filenames)
sorted_files = yield _SortChunksPipeline(job_name, hashed_files)
merged_files = yield _MergePipeline(job_name, sorted_files)
with pipeline.After(merged_files):
all_temp_files = yield pipeline_common.Extend(
hashed_files, sorted_files)
yield mapper_pipeline._CleanupPipeline(all_temp_files)
yield pipeline_common.Return(merged_files)
def run(self, job_name, filenames):
sort_mappers = []
for i in range(len(filenames)):
filename = filenames[i]
sort_mapper = yield mapper_pipeline.MapperPipeline(
"%s-shuffle-sort-%s" % (job_name, str(i)),
__name__ + "._sort_records_map",
__name__ + "._BatchRecordsReader",
None, {
"files": [filename],
"processing_rate": 1000000,
},
shards=1)
sort_mappers.append(sort_mapper)
with pipeline.After(*sort_mappers):
job_ids = yield pipeline_common.Append(
*[mapper.job_id for mapper in sort_mappers])
result = yield _CollectOutputFiles(job_ids)
with pipeline.After(result):
yield _CleanupOutputFiles(job_ids)
yield pipeline_common.Return(result)
def run(self, filenames):
mapper = yield mapper_pipeline.MapperPipeline(
"sort",
__name__ + "._sort_records",
__name__ + "._BatchRecordsReader",
None,
{
"files": filenames,
"processing_rate": 1000000,
},
shards=1)
with pipeline.After(mapper):
yield _CollectOutputFiles(mapper.job_id)
def run(self, job_name, filenames, shards=None):
if files.shuffler.available():
yield _ShuffleServicePipeline(job_name, filenames)
else:
hashed_files = yield _HashPipeline(job_name, filenames, shards=shards)
sorted_files = yield _SortChunksPipeline(job_name, hashed_files)
temp_files = [hashed_files, sorted_files]
merged_files = yield _MergePipeline(job_name, sorted_files)
with pipeline.After(merged_files):
all_temp_files = yield pipeline_common.Extend(*temp_files)
yield mapper_pipeline._CleanupPipeline(all_temp_files)
yield pipeline_common.Return(merged_files)
def run(self, job_name, mapper_params, filenames, shards=None):
bucket_name = mapper_params["bucket_name"]
hashed_files = yield _HashPipeline(job_name, bucket_name,
filenames, shards=shards)
sorted_files = yield _SortChunksPipeline(job_name, bucket_name,
hashed_files)
temp_files = [hashed_files, sorted_files]
merged_files = yield _MergePipeline(job_name, bucket_name, sorted_files)
with pipeline.After(merged_files):
all_temp_files = yield pipeline_common.Extend(*temp_files)
yield _GCSCleanupPipeline(all_temp_files)
yield pipeline_common.Return(merged_files)
def run(self,
job_name,
mapper_spec,
reducer_spec,
input_reader_spec,
output_writer_spec=None,
mapper_params=None,
reducer_params=None,
shards=None,
combiner_spec=None):
if mapper_params.get("bucket_name") is None:
try:
mapper_params["bucket_name"] = (
app_identity.get_default_gcs_bucket_name())
except Exception as e:
raise errors.Error("Unable to get the GCS default bucket name. "
"Check to see that GCS is properly activated. "
+ str(e))
if mapper_params["bucket_name"] is None:
raise errors.Error("There is no GCS default bucket name. "
"Check to see that GCS is properly activated.")
map_pipeline = yield MapPipeline(job_name,
mapper_spec,
input_reader_spec,
params=mapper_params,
shards=shards)
shuffler_pipeline = yield ShufflePipeline(
job_name, mapper_params, map_pipeline)
reducer_pipeline = yield ReducePipeline(
job_name,
reducer_spec,
output_writer_spec,
reducer_params,
mapper_params["bucket_name"],
shuffler_pipeline,
combiner_spec=combiner_spec)
with pipeline.After(reducer_pipeline):
all_temp_files = yield pipeline_common.Extend(
map_pipeline, shuffler_pipeline)
yield CleanupPipeline(all_temp_files)
yield _ReturnPipeline(map_pipeline.result_status,
reducer_pipeline.result_status,
reducer_pipeline)
class MapreducePipeline(pipeline_base._OutputSlotsMixin,
pipeline_base.PipelineBase):
"""Pipeline to execute MapReduce jobs.
The Shuffle stage uses Google Cloud Storage (GCS). For newly created projects,
GCS is activated automatically. To activate GCS follow these instructions:
https://cloud.google.com/storage/docs/signup#activate
Args:
job_name: job name as string.
mapper_spec: specification of mapper to use.
reducer_spec: specification of reducer to use.
input_reader_spec: specification of input reader to read data from.
output_writer_spec: specification of output writer to save reduce output to.
mapper_params: parameters to use for mapper phase.
reducer_params: parameters to use for reduce phase.
shards: number of shards to use as int.
combiner_spec: Optional. Specification of a combine function. If not
supplied, no combine step will take place. The combine function takes a
key, list of values and list of previously combined results. It yields
combined values that might be processed by another combiner call, but will
eventually end up in reducer. The combiner output key is assumed to be the
same as the input key.
Returns:
result_status: one of model.MapreduceState._RESULTS. Check this to see
if the job is successful.
default: a list of filenames if the mapreduce was successful and
was outputting files. An empty list otherwise.
"""
def run(self,
job_name,
mapper_spec,
reducer_spec,
input_reader_spec,
output_writer_spec=None,
mapper_params=None,
reducer_params=None,
shards=None,
combiner_spec=None):
if mapper_params.get("bucket_name") is None:
try:
mapper_params["bucket_name"] = (
app_identity.get_default_gcs_bucket_name())
except Exception, e:
raise errors.Error(
"Unable to get the GCS default bucket name. "
"Check to see that GCS is properly activated. " + str(e))
if mapper_params["bucket_name"] is None:
raise errors.Error("There is no GCS default bucket name. "
"Check to see that GCS is properly activated.")
map_pipeline = yield MapPipeline(job_name,
mapper_spec,
input_reader_spec,
params=mapper_params,
shards=shards)
shuffler_pipeline = yield ShufflePipeline(job_name, mapper_params,
map_pipeline)
reducer_pipeline = yield ReducePipeline(job_name,
reducer_spec,
output_writer_spec,
reducer_params,
shuffler_pipeline,
combiner_spec=combiner_spec)
with pipeline.After(reducer_pipeline):
all_temp_files = yield pipeline_common.Extend(
map_pipeline, shuffler_pipeline)
yield CleanupPipeline(all_temp_files)
yield _ReturnPipeline(map_pipeline.result_status,
reducer_pipeline.result_status, reducer_pipeline)
