如何在 Python 中创建从 Pub/Sub 到 GCS 的数据流管道 [英] How to create a Dataflow pipeline from Pub/Sub to GCS in Python

查看：24
本文介绍了如何在 Python 中创建从 Pub/Sub 到 GCS 的数据流管道的处理方法，对大家解决问题具有一定的参考价值，需要的朋友们下面随着小编来一起学习吧！

问题描述

我想使用 Dataflow 将数据从 Pub/Sub 移动到 GCS.所以基本上我希望 Dataflow 在固定的时间内(例如 15 分钟)积累一些消息，然后在这段时间过去后将这些数据作为文本文件写入 GCS.

I want to use Dataflow to move data from Pub/Sub to GCS. So basically I want Dataflow to accumulate some messages in a fixed amount of time (15 minutes for example), then write those data as text file into GCS when that amount of time has passed.

我的最终目标是创建一个自定义管道，所以Pub/Sub 到 Cloud Storage"模板对我来说还不够，而且我完全不了解 Java，这让我开始在 Python 中进行调整.

My final goal is to create a custom pipeline, so "Pub/Sub to Cloud Storage" template is not enough for me, and I don't know about Java at all, which made me start to tweak in Python.

这是我目前所拥有的(Apache Beam Python SDK 2.10.0):

Here is what I have got as of now (Apache Beam Python SDK 2.10.0):

import apache_beam as beam

TOPIC_PATH="projects/<my-project>/topics/<my-topic>"

def CombineFn(e):
    return "\n".join(e)

o = beam.options.pipeline_options.PipelineOptions()
p = beam.Pipeline(options=o)
data = ( p | "Read From Pub/Sub" >> beam.io.ReadFromPubSub(topic=TOPIC_PATH)
       | "Window" >> beam.WindowInto(beam.window.FixedWindows(30))
       | "Combine" >> beam.transforms.core.CombineGlobally(CombineFn).without_defaults()
       | "Output" >> beam.io.WriteToText("<GCS path or local path>"))

res = p.run()
res.wait_until_finish()

我在本地环境中运行这个程序没有问题.

I ran this program without problems in local environment.

python main.py

它在本地运行，但从指定的 Pub/Sub 主题读取，并按预期每 30 秒过去后写入指定的 GCS 路径.

It runs locally but read from specified Pub/Sub topic and writes to the specified GCS path every time the 30 seconds has passed, as expected.

然而，现在的问题是，当我在 Google Cloud Platform(即 Cloud Dataflow)上运行它时，它不断发出神秘的异常.

The problem now is, however, when I run this on the Google Cloud Platform, namely Cloud Dataflow, it continuously emits mysterious Exception.

java.util.concurrent.ExecutionException: java.lang.RuntimeException: Error received from SDK harness for instruction -1096: Traceback (most recent call last):
  File "/usr/local/lib/python2.7/dist-packages/apache_beam/runners/worker/sdk_worker.py", line 148, in _execute
    response = task()
  File "/usr/local/lib/python2.7/dist-packages/apache_beam/runners/worker/sdk_worker.py", line 183, in <lambda>
    self._execute(lambda: worker.do_instruction(work), work)
  File "/usr/local/lib/python2.7/dist-packages/apache_beam/runners/worker/sdk_worker.py", line 256, in do_instruction
    request.instruction_id)
  File "/usr/local/lib/python2.7/dist-packages/apache_beam/runners/worker/sdk_worker.py", line 272, in process_bundle
    bundle_processor.process_bundle(instruction_id)
  File "/usr/local/lib/python2.7/dist-packages/apache_beam/runners/worker/bundle_processor.py", line 494, in process_bundle
    op.finish()
  File "apache_beam/runners/worker/operations.py", line 506, in apache_beam.runners.worker.operations.DoOperation.finish
    def finish(self):
  File "apache_beam/runners/worker/operations.py", line 507, in apache_beam.runners.worker.operations.DoOperation.finish
    with self.scoped_finish_state:
  File "apache_beam/runners/worker/operations.py", line 508, in apache_beam.runners.worker.operations.DoOperation.finish
    self.dofn_runner.finish()
  File "apache_beam/runners/common.py", line 703, in apache_beam.runners.common.DoFnRunner.finish
    self._invoke_bundle_method(self.do_fn_invoker.invoke_finish_bundle)
  File "apache_beam/runners/common.py", line 697, in apache_beam.runners.common.DoFnRunner._invoke_bundle_method
    self._reraise_augmented(exn)
  File "apache_beam/runners/common.py", line 722, in apache_beam.runners.common.DoFnRunner._reraise_augmented
    raise_with_traceback(new_exn)
  File "apache_beam/runners/common.py", line 695, in apache_beam.runners.common.DoFnRunner._invoke_bundle_method
    bundle_method()
  File "apache_beam/runners/common.py", line 361, in apache_beam.runners.common.DoFnInvoker.invoke_finish_bundle
    def invoke_finish_bundle(self):
  File "apache_beam/runners/common.py", line 364, in apache_beam.runners.common.DoFnInvoker.invoke_finish_bundle
    self.output_processor.finish_bundle_outputs(
  File "apache_beam/runners/common.py", line 832, in apache_beam.runners.common._OutputProcessor.finish_bundle_outputs
    self.main_receivers.receive(windowed_value)
  File "apache_beam/runners/worker/operations.py", line 87, in apache_beam.runners.worker.operations.ConsumerSet.receive
    self.update_counters_start(windowed_value)
  File "apache_beam/runners/worker/operations.py", line 93, in apache_beam.runners.worker.operations.ConsumerSet.update_counters_start
    self.opcounter.update_from(windowed_value)
  File "apache_beam/runners/worker/opcounters.py", line 195, in apache_beam.runners.worker.opcounters.OperationCounters.update_from
    self.do_sample(windowed_value)
  File "apache_beam/runners/worker/opcounters.py", line 213, in apache_beam.runners.worker.opcounters.OperationCounters.do_sample
    self.coder_impl.get_estimated_size_and_observables(windowed_value))
  File "apache_beam/coders/coder_impl.py", line 953, in apache_beam.coders.coder_impl.WindowedValueCoderImpl.get_estimated_size_and_observables
    def get_estimated_size_and_observables(self, value, nested=False):
  File "apache_beam/coders/coder_impl.py", line 969, in apache_beam.coders.coder_impl.WindowedValueCoderImpl.get_estimated_size_and_observables
    self._windows_coder.estimate_size(value.windows, nested=True))
  File "apache_beam/coders/coder_impl.py", line 758, in apache_beam.coders.coder_impl.SequenceCoderImpl.estimate_size
    self.get_estimated_size_and_observables(value))
  File "apache_beam/coders/coder_impl.py", line 772, in apache_beam.coders.coder_impl.SequenceCoderImpl.get_estimated_size_and_observables
    self._elem_coder.get_estimated_size_and_observables(
  File "apache_beam/coders/coder_impl.py", line 134, in apache_beam.coders.coder_impl.CoderImpl.get_estimated_size_and_observables
    return self.estimate_size(value, nested), []
  File "apache_beam/coders/coder_impl.py", line 458, in apache_beam.coders.coder_impl.IntervalWindowCoderImpl.estimate_size
    typed_value = value
TypeError: Cannot convert GlobalWindow to apache_beam.utils.windowed_value._IntervalWindowBase [while running 'generatedPtransform-1090']

        java.util.concurrent.CompletableFuture.reportGet(CompletableFuture.java:357)
        java.util.concurrent.CompletableFuture.get(CompletableFuture.java:1895)
        org.apache.beam.sdk.util.MoreFutures.get(MoreFutures.java:57)
        org.apache.beam.runners.dataflow.worker.fn.control.RegisterAndProcessBundleOperation.finish(RegisterAndProcessBundleOperation.java:280)
        org.apache.beam.runners.dataflow.worker.util.common.worker.MapTaskExecutor.execute(MapTaskExecutor.java:84)
        org.apache.beam.runners.dataflow.worker.fn.control.BeamFnMapTaskExecutor.execute(BeamFnMapTaskExecutor.java:130)
        org.apache.beam.runners.dataflow.worker.StreamingDataflowWorker.process(StreamingDataflowWorker.java:1233)
        org.apache.beam.runners.dataflow.worker.StreamingDataflowWorker.access$1000(StreamingDataflowWorker.java:144)
        org.apache.beam.runners.dataflow.worker.StreamingDataflowWorker$6.run(StreamingDataflowWorker.java:972)
        java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
        java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
        java.lang.Thread.run(Thread.java:745)
Caused by: java.lang.RuntimeException: Error received from SDK harness for instruction -1096: Traceback (most recent call last):
  File "/usr/local/lib/python2.7/dist-packages/apache_beam/runners/worker/sdk_worker.py", line 148, in _execute
    response = task()
  File "/usr/local/lib/python2.7/dist-packages/apache_beam/runners/worker/sdk_worker.py", line 183, in <lambda>
    self._execute(lambda: worker.do_instruction(work), work)
  File "/usr/local/lib/python2.7/dist-packages/apache_beam/runners/worker/sdk_worker.py", line 256, in do_instruction
    request.instruction_id)
  File "/usr/local/lib/python2.7/dist-packages/apache_beam/runners/worker/sdk_worker.py", line 272, in process_bundle
    bundle_processor.process_bundle(instruction_id)
  File "/usr/local/lib/python2.7/dist-packages/apache_beam/runners/worker/bundle_processor.py", line 494, in process_bundle
    op.finish()
  File "apache_beam/runners/worker/operations.py", line 506, in apache_beam.runners.worker.operations.DoOperation.finish
    def finish(self):
  File "apache_beam/runners/worker/operations.py", line 507, in apache_beam.runners.worker.operations.DoOperation.finish
    with self.scoped_finish_state:
  File "apache_beam/runners/worker/operations.py", line 508, in apache_beam.runners.worker.operations.DoOperation.finish
    self.dofn_runner.finish()
  File "apache_beam/runners/common.py", line 703, in apache_beam.runners.common.DoFnRunner.finish
    self._invoke_bundle_method(self.do_fn_invoker.invoke_finish_bundle)
  File "apache_beam/runners/common.py", line 697, in apache_beam.runners.common.DoFnRunner._invoke_bundle_method
    self._reraise_augmented(exn)
  File "apache_beam/runners/common.py", line 722, in apache_beam.runners.common.DoFnRunner._reraise_augmented
    raise_with_traceback(new_exn)
  File "apache_beam/runners/common.py", line 695, in apache_beam.runners.common.DoFnRunner._invoke_bundle_method
    bundle_method()
  File "apache_beam/runners/common.py", line 361, in apache_beam.runners.common.DoFnInvoker.invoke_finish_bundle
    def invoke_finish_bundle(self):
  File "apache_beam/runners/common.py", line 364, in apache_beam.runners.common.DoFnInvoker.invoke_finish_bundle
    self.output_processor.finish_bundle_outputs(
  File "apache_beam/runners/common.py", line 832, in apache_beam.runners.common._OutputProcessor.finish_bundle_outputs
    self.main_receivers.receive(windowed_value)
  File "apache_beam/runners/worker/operations.py", line 87, in apache_beam.runners.worker.operations.ConsumerSet.receive
    self.update_counters_start(windowed_value)
  File "apache_beam/runners/worker/operations.py", line 93, in apache_beam.runners.worker.operations.ConsumerSet.update_counters_start
    self.opcounter.update_from(windowed_value)
  File "apache_beam/runners/worker/opcounters.py", line 195, in apache_beam.runners.worker.opcounters.OperationCounters.update_from
    self.do_sample(windowed_value)
  File "apache_beam/runners/worker/opcounters.py", line 213, in apache_beam.runners.worker.opcounters.OperationCounters.do_sample
    self.coder_impl.get_estimated_size_and_observables(windowed_value))
  File "apache_beam/coders/coder_impl.py", line 953, in apache_beam.coders.coder_impl.WindowedValueCoderImpl.get_estimated_size_and_observables
    def get_estimated_size_and_observables(self, value, nested=False):
  File "apache_beam/coders/coder_impl.py", line 969, in apache_beam.coders.coder_impl.WindowedValueCoderImpl.get_estimated_size_and_observables
    self._windows_coder.estimate_size(value.windows, nested=True))
  File "apache_beam/coders/coder_impl.py", line 758, in apache_beam.coders.coder_impl.SequenceCoderImpl.estimate_size
    self.get_estimated_size_and_observables(value))
  File "apache_beam/coders/coder_impl.py", line 772, in apache_beam.coders.coder_impl.SequenceCoderImpl.get_estimated_size_and_observables
    self._elem_coder.get_estimated_size_and_observables(
  File "apache_beam/coders/coder_impl.py", line 134, in apache_beam.coders.coder_impl.CoderImpl.get_estimated_size_and_observables
    return self.estimate_size(value, nested), []
  File "apache_beam/coders/coder_impl.py", line 458, in apache_beam.coders.coder_impl.IntervalWindowCoderImpl.estimate_size
    typed_value = value
TypeError: Cannot convert GlobalWindow to apache_beam.utils.windowed_value._IntervalWindowBase [while running 'generatedPtransform-1090']

        org.apache.beam.runners.fnexecution.control.FnApiControlClient$ResponseStreamObserver.onNext(FnApiControlClient.java:157)
        org.apache.beam.runners.fnexecution.control.FnApiControlClient$ResponseStreamObserver.onNext(FnApiControlClient.java:140)
        org.apache.beam.vendor.grpc.v1p13p1.io.grpc.stub.ServerCalls$StreamingServerCallHandler$StreamingServerCallListener.onMessage(ServerCalls.java:248)
        org.apache.beam.vendor.grpc.v1p13p1.io.grpc.ForwardingServerCallListener.onMessage(ForwardingServerCallListener.java:33)
        org.apache.beam.vendor.grpc.v1p13p1.io.grpc.Contexts$ContextualizedServerCallListener.onMessage(Contexts.java:76)
        org.apache.beam.vendor.grpc.v1p13p1.io.grpc.internal.ServerCallImpl$ServerStreamListenerImpl.messagesAvailable(ServerCallImpl.java:263)
        org.apache.beam.vendor.grpc.v1p13p1.io.grpc.internal.ServerImpl$JumpToApplicationThreadServerStreamListener$1MessagesAvailable.runInContext(ServerImpl.java:683)
        org.apache.beam.vendor.grpc.v1p13p1.io.grpc.internal.ContextRunnable.run(ContextRunnable.java:37)
        org.apache.beam.vendor.grpc.v1p13p1.io.grpc.internal.SerializingExecutor.run(SerializingExecutor.java:123)
        java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
        java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
        java.lang.Thread.run(Thread.java:745)

每次尝试写入 GCS 时，都会以非阻塞方式显示上述异常.这导致我的情况是，当它尝试输出时，肯定会生成一个新的文本文件，但文本内容始终与第一个窗口输出相同.这显然是不需要的.

Every time it tries to write to GCS, the exception above is shown without in a non-blocking way. Which leads me to a situation that, when it attempts to output, a new text file is certainly generated but the text content is always the same as the first windowed output. This is obviously unwanted.

异常在堆栈跟踪中嵌套如此之深，以至于很难猜测根本原因是什么，而且我不知道为什么它在 DirectRunner 上运行良好，但在 DataflowRunner 上却完全没有.似乎在管道的某个地方，全局窗口值被转换为非全局窗口值，尽管我在管道的第二阶段使用了非全局窗口转换.添加自定义触发器没有帮助.

The exception is so deeply nested in the stack trace that it is extremely hard to guess what the root cause is, and I have no idea why it ran fine on DirectRunner but not at all on DataflowRunner. It seems it says somewhere in the pipeline, global windowed values are converted to non-global windowed values, although I used non-global window transform at the second stage of the pipeline. Adding custom triggers didn't help.

推荐答案

我遇到了同样的错误，并找到了解决方法，但没有解决:

I ran into this same error, and found a workaround, but not a fix:

TypeError: Cannot convert GlobalWindow to apache_beam.utils.windowed_value._IntervalWindowBase [while running 'test-file-out/Write/WriteImpl/WriteBundles']

使用 DirectRunner 在本地运行，使用 DataflowRunner 在数据流上运行.

running locally with DirectRunner and on dataflow with DataflowRunner.

恢复到 apache-beam[gcp]==2.9.0 允许我的管道按预期运行.

Reverting to apache-beam[gcp]==2.9.0 allows my pipeline to run as expected.

这篇关于如何在 Python 中创建从 Pub/Sub 到 GCS 的数据流管道的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！

查看全文
相关文章
Python最新文章
热门教程
热门工具
登录 关闭
扫码关注1秒登录
发送“验证码”获取 | 15天全站免登陆