Configure and troubleshoot Apache Beam runners, the engines that execute data processing pipelines. Covers local testing, Dataflow, Flink, and Spark, with Java and Python examples, runner validation tests, and guidance on logs and execution settings.
中文简介正在准备中。当前显示英语简介。
Guides understanding and working with Apache Beam runners (Direct, Dataflow, Flink, Spark, etc.). Use when configuring pipelines for different execution environments or debugging runner-specific issues.
安装前可以查看交给 AI 代理的指令内容。
Runners execute Beam pipelines on distributed processing backends. Each runner translates the portable Beam model to its native execution engine.
| Runner | Location | Description |
|---|---|---|
| Direct | runners/direct-java/ | Local execution for testing |
| Prism | runners/prism/ | Portable local runner |
| Dataflow | runners/google-cloud-dataflow-java/ | Google Cloud Dataflow |
| Flink | runners/flink/ | Apache Flink |
| Spark | runners/spark/ | Apache Spark |
| Jet | runners/jet/ | Hazelcast Jet |
| Twister2 | runners/twister2/ | Twister2 |
For local development and testing.
PipelineOptions options = PipelineOptionsFactory.create();
options.setRunner(DirectRunner.class);
Pipeline p = Pipeline.create(options);
options = PipelineOptions()
options.view_as(StandardOptions).runner = 'DirectRunner'
p = beam.Pipeline(options=options)
--runner=DirectRunner
DataflowPipelineOptions options = PipelineOptionsFactory.as(DataflowPipelineOptions.class);
options.setRunner(DataflowRunner.class);
options.setProject("my-project");
options.setRegion("us-central1");
options.setTempLocation("gs://my-bucket/temp");
options = PipelineOptions([
'--runner=DataflowRunner',
'--project=my-project',
'--region=us-central1',
'--temp_location=gs://my-bucket/temp'
])
--experiments=use_runner_v2
--sdkContainerImage=gcr.io/project/beam_java11_sdk:custom
FlinkPipelineOptions options = PipelineOptionsFactory.as(FlinkPipelineOptions.class);
options.setRunner(FlinkRunner.class);
options.setFlinkMaster("[local]");
options.setFlinkMaster("host:port");
options = PipelineOptions([
'--runner=FlinkRunner',
'--flink_master=host:port',
'--environment_type=LOOPBACK' # or DOCKER, EXTERNAL
])
SparkPipelineOptions options = PipelineOptionsFactory.as(SparkPipelineOptions.class);
options.setRunner(SparkRunner.class);
options.setSparkMaster("local[*]"); # or spark://host:port
options = PipelineOptions([
'--runner=SparkRunner',
'--spark_master_url=local[*]'
])
Tests that validate runner correctness:
# Direct Runner
./gradlew :runners:direct-java:validatesRunner
# Flink Runner
./gradlew :runners:flink:1.18:validatesRunner
# Spark Runner
./gradlew :runners:spark:3:validatesRunner
# Dataflow Runner
./gradlew :runners:google-cloud-dataflow-java:validatesRunner
@Rule public TestPipeline pipeline = TestPipeline.create();
// Set runner via system property
-DbeamTestPipelineOptions='["--runner=TestDataflowRunner"]'
DOCKER - SDK in Docker containerLOOPBACK - SDK in same process (testing)EXTERNAL - SDK at specified addressPROCESS - SDK in subprocessStart Flink job server:
./gradlew :runners:flink:1.18:job-server:runShadow
Start Spark job server:
./gradlew :runners:spark:3:job-server:runShadow
| Option | Description |
|---|---|
--project | GCP project |
--region | GCP region |
--tempLocation | GCS temp location |
--stagingLocation | GCS staging |
--numWorkers | Initial workers |
--maxNumWorkers | Max workers |
--workerMachineType | VM type |
| Option | Description |
|---|---|
--flinkMaster | Flink master address |
--parallelism | Default parallelism |
--checkpointingInterval | Checkpoint interval |
| Option | Description |
|---|---|
--sparkMaster | Spark master URL |
--sparkConf | Additional Spark config |
./gradlew :runners:google-cloud-dataflow-java:worker:shadowJar
./gradlew :runners:flink:1.18:job-server:shadowJar
./gradlew :runners:spark:3:job-server:shadowJar
-Dorg.slf4j.simpleLogger.defaultLogLevel=debug--targetParallelism=1 for deterministic execution--experiments=upload_graph for graph debugging还没有评价。欢迎分享你的使用感受。
无需登录即可发表。欢迎分享使用感受、提问，或说明在哪些环境下无法运行。
还没有评论。来写第一条吧。
Guides adding metadata, extra information attached to data elements, to Apache Beam's WindowedValue. Maps the changes needed across serialization, storage, processing, and runner interfaces so new fields remain available during pipeline execution.
中文简介正在准备中。当前显示英语简介。
Learn Apache Beam's core concepts through Java and Python examples. Understand distributed datasets, processing steps, pipeline execution, and time windows, with guidance on aggregation, late data, error handling, and sharing processing steps across languages.
中文简介正在准备中。当前显示英语简介。
Helps investigate and modify Apache Beam's automated build and test workflows in GitHub Actions. Explains workflow types, failure logs, targeted reruns, local reproduction, and the cloud credentials needed for workflows that access GCP.
中文简介正在准备中。当前显示英语简介。
Navigate Apache Beam contributions from finding an issue to submitting a pull request, a proposed code change, and responding to reviews. Covers environment setup, tests, formatting, licensing requirements, and closer evaluation of changes to core components.
中文简介正在准备中。当前显示英语简介。
Build Apache Beam connectors that read and write external data. This guide covers Java implementation, schema support for structured records, access from Python and Go, Managed API integration, and testing.
中文简介正在准备中。当前显示英语简介。
Build and test Apache Beam with Gradle, the system that manages its project tasks. Find relevant configuration files, inspect library dependencies, run formatting checks, publish locally, and investigate build failures.
中文简介正在准备中。当前显示英语简介。