Does GCP Dataflow support kafka IO in python?

Question

I am trying to read data from kafka topic using kafka.ReadFromKafka() method in python code.My code looks like below:

from apache_beam.io.external import kafka
import apache_beam as beam

options = PipelineOptions()

with beam.Pipeline(options=options) as p:
           plants = (
      p
        |       'read' >> kafka.ReadFromKafka({'bootstrap.servers': 'public_ip:9092'}, ['topic1']))

But getting below error message.

ERROR:apache_beam.runners.runner:Error while visiting read Traceback (most recent call last): File "test_file.py", line 16, in <module> | 'read' >> kafka.ReadFromKafka({'bootstrap.servers': 'localhost:9092'}, ['topic1']) File "/usr/local/lib/python3.7/dist-packages/apache_beam/pipeline.py", line 547, in __exit__ self.run().wait_until_finish() File "/usr/local/lib/python3.7/dist-packages/apache_beam/pipeline.py", line 526, in run return self.runner.run_pipeline(self, self._options) File "/usr/local/lib/python3.7/dist-packages/apache_beam/runners/dataflow/dataflow_runner.py", line 565, in run_pipeline self.visit_transforms(pipeline, options) File "/usr/local/lib/python3.7/dist-packages/apache_beam/runners/runner.py", line 224, in visit_transforms pipeline.visit(RunVisitor(self)) File "/usr/local/lib/python3.7/dist-packages/apache_beam/pipeline.py", line 572, in visit self._root_transform().visit(visitor, self, visited) File "/usr/local/lib/python3.7/dist-packages/apache_beam/pipeline.py", line 1075, in visit part.visit(visitor, pipeline, visited) File "/usr/local/lib/python3.7/dist-packages/apache_beam/pipeline.py", line 1078, in visit visitor.visit_transform(self) File "/usr/local/lib/python3.7/dist-packages/apache_beam/runners/runner.py", line 219, in visit_transform self.runner.run_transform(transform_node, options) File "/usr/local/lib/python3.7/dist-packages/apache_beam/runners/runner.py", line 249, in run_transform (transform_node.transform, self)) NotImplementedError: Execution of [<ReadFromKafka(PTransform) label=[ReadFromKafka(beam:external:java:kafka:read:v1)]>] not implemented in runner <apache_beam.runners.dataflow.dataflow_runner.DataflowRunner object at 0x7f72463344a8>.

Is it because apache beam Dataflow runner doesn't support kafkaIO ?

You are running as DataflowRunner and connecting to localhost:9092? — bigbounty
i have added public ip in "advertised.listeners" in server.properties file as well as in code. Still facing same error. — Joseph N
But Dataflow can't access your localhost if you are running in google servers — bigbounty
Any suggesions, whats should i do in this situation? i have my kafka in gcp compute instance — Joseph N
If you want to test your pipeline, you can run your pipeline using DirectRunner on your local machine. If you want the code on direct runner, I can write it as an answer — bigbounty

Jayadeep Jayaraman Jayadeep Jayaraman · Accepted Answer · 2020-07-07T13:40:37

The python SDK for beam does support connecting to Kafka. Below is a code snippet

from __future__ import print_function
import apache_beam as beam
from apache_beam.options.pipeline_options import PipelineOptions
from beam_nuggets.io import kafkaio

kafka_topic = "notifications"
kafka_config = {"topic": kafka_topic,
                "bootstrap_servers": "localhost:9092",
                "group_id": "notification_consumer_group"}

with beam.Pipeline(options=PipelineOptions()) as p:
    notifications = p | "Reading messages from Kafka" >> kafkaio.KafkaConsume(kafka_config)
    notifications | 'Writing to stdout' >> beam.Map(print)

The bootstrap_servers is a comma separated host and port configuration where your brokers are deployed. You will get this information from your Kafka cluster configuration.

Does GCP Dataflow support kafka IO in python?

2 Answers