3

I have this very simplified use case: I want to use Apache Flink (1.11) to read data from a Kafka topic (let's call it source_topic), count an attribute in it (called b) and write the result into another Kafka topic (result_topic).

I have the following code so far:

from pyflink.datastream import StreamExecutionEnvironment, TimeCharacteristic
from pyflink.table import StreamTableEnvironment, EnvironmentSettings

def log_processing():
    env = StreamExecutionEnvironment.get_execution_environment()
    env_settings = EnvironmentSettings.new_instance().use_blink_planner().in_streaming_mode().build()
    t_env = StreamTableEnvironment.create(stream_execution_environment=env, environment_settings=env_settings)`
    t_env.get_config().get_configuration().set_boolean("python.fn-execution.memory.managed", True)
    t_env.get_config().get_configuration().set_string("pipeline.jars", "file:///opt/flink-1.11.2/lib/flink-sql-connector-kafka_2.11-1.11.2.jar")

    source_ddl = """
            CREATE TABLE source_table(
                a STRING,
                b INT
            ) WITH (
              'connector' = 'kafka',
              'topic' = 'source_topic',
              'properties.bootstrap.servers' = 'node-1:9092',
              'scan.startup.mode' = 'earliest-offset',
              'format' = 'csv',
              'csv.ignore-parse-errors' = 'true'
            )
            """

    sink_ddl = """
            CREATE TABLE result_table(
                b INT,
                result BIGINT
            ) WITH (
              'connector' = 'kafka',
              'topic' = 'result_topic',
              'properties.bootstrap.servers' = 'node-1:9092',
              'format' = 'csv'
            )
            """

    t_env.execute_sql(source_ddl)
    t_env.execute_sql(sink_ddl)
    t_env.execute_sql("INSERT INTO result_table SELECT b,COUNT(b) FROM source_table GROUP BY b")
    t_env.execute("Kafka_Flink_job")

if __name__ == '__main__':
    log_processing()

But when I execute it, I get the following error:

py4j.protocol.Py4JJavaError: An error occurred while calling o5.executeSql.
: org.apache.flink.table.api.TableException: Table sink 'default_catalog.default_database.result_table' doesn't support consuming update changes which is produced by node GroupAggregate(groupBy=[b], select=[b, COUNT(b) AS EXPR$1])

I am able to write data into a Kafka topic with a simple SELECT statement. But as soon as I add the GROUP BY clause, the exception above is thrown. I followed Flink's documentation on the use of the Table API with SQL for Python: https://ci.apache.org/projects/flink/flink-docs-release-1.11/dev/table/common.html#sql

Any help is highly appreciated, I am very new to Stream Processing and Flink. Thank you!

Irina S.
  • 133
  • 8

1 Answers1

2

Using a GROUP BY clause will generate an updating stream, which is not supported by the Kafka connector as of Flink 1.11. On the other hand, when you use a simple SELECT statement without any aggregation, the result stream is append-only (this is why you're able to consume it without issues).

Flink 1.12 is very close to being released, and it includes a new upsert Kafka connector (FLIP-149, if you're curious) that will allow you to do this type of operation also in PyFlink (i.e. the Python Table API).

morsapaes
  • 436
  • 2
  • 7
  • Thanks a lot for your answer, morsapaes. So, I will wait for Flink 1.12 which is planned to be released at the end of November to use a Kafka sink for updated data. – Irina S. Nov 17 '20 at 09:51