Beam Streaming 管道不会将文件写入存储桶

UI 在 GCP Dataflow 上有一个 python 流管道，可以从 PubSub 读取数千条消息，如下所示：

    with beam.Pipeline(options=pipeline_options) as p:
      lines = p | "read" >> ReadFromPubSub(topic=str(job_options.inputTopic))
      lines = lines | "decode" >> beam.Map(decode_message)
      lines = lines | "Parse" >> beam.Map(parse_json)
      lines = lines | beam.WindowInto(beam.window.FixedWindows(1*60))
      lines = lines | "Add device id key" >> beam.Map(lambda elem: (elem.get('id'), elem))
      lines = lines | "Group by key" >> beam.GroupByKey()
      lines = lines | "Abandon key" >> beam.Map(flatten)
      lines | "WriteToAvro" >> beam.io.WriteToAvro(job_options.outputLocation, schema=schema, file_name_suffix='.avro', mime_type='application/x-avro')

管道运行得很好，只是它从不产生任何输出。有什么想法吗？

您的代码似乎存在一些问题。首先，有一些关于 null/None（您已经修复）和 ints/floats（在注释中指出）的格式错误的数据。最后，写入Avro转换无法写入无限的 PCollection。有一个解决方法，您可以在其中定义一个新的sink并将其与写入文件能够写入无界 PCollection 的转换。

请注意，截至撰写本文时 (2020-06-18)，此方法不适用于 Apache Beam Python SDK BEAM-6522）。在这种情况下，这会强制解决方案改用 FastAvro。如果手动升级 dill 则可以使用 Avro 至 >= 0.3.1.1andAvro >= 1.9.0，但要小心，因为目前尚未测试。

解决方法如下：

from apache_beam.io.fileio import FileSink
from apache_beam.io.fileio import WriteToFiles
import fastavro

class AvroFileSink(FileSink):
    def __init__(self, schema, codec='deflate'):
        self._schema = schema
        self._codec = codec

    def open(self, fh):
        # This is called on every new bundle.
        self.writer = fastavro.write.Writer(fh, self._schema, self._codec)

    def write(self, record):
        # This is called on every element.
        self.writer.write(record)

    def flush(self):
        self.writer.flush()

这个新水槽的使用方式如下：

import apache_beam as beam

# Replace the following with your schema.
schema = fastavro.schema.parse_schema({
    'name': 'row',
    'namespace': 'test',
    'type': 'record',
    'fields': [
        {'name': 'a', 'type': 'int'},
    ],
})

# Create the sink. This will be used by the WriteToFiles transform to write
# individual elements to the Avro file.
sink = AvroFileSink(schema=schema)

with beam.Pipeline(...) as p:
    lines = p | beam.ReadFromPubSub(...)
    lines = ...

    # This is where your new sink gets used. The WriteToFiles transform takes
    # the sink and uses it to write to a directory defined by the path 
    # argument.
    lines | WriteToFiles(path=job_options.outputLocation, sink=sink)