type: runners title: “Twister2 Runner” aliases: /learn/runners/twister2/

Overview

Note Twister2 runner is deprecated and the support is planned to be removed in Beam 3.0 (Issue).

Twister2 Runner can be used to execute Apache Beam pipelines on top of a Twister2 cluster. Twister2 Runner runs Beam pipelines as Twister2 jobs, which can be executed on a Twister2 cluster either as a local deployment or distributed deployment using, Nomad, Kubernetes, Slurm, etc.

The Twister2 runner is suitable for large scale batch jobs, specially jobs that require high performance, and provide.

  • Batch pipeline support.
  • Support for HPC environments, supports propriety interconnects such as Infiniband.
  • Distributed massively parallel data processing engine with high performance using Bulk Synchronous Parallel (BSP) style execution.
  • Native support for Beam side-inputs.

The Beam Capability Matrix documents the supported capabilities of the Twister2 Runner.

Running WordCount with the Twister2 Runner

Generating the Beam examples project

Just follow the instruction from the Java Quickstart page

Running WordCount on a Twister2 Local Deployment

Issue following command in the Beam examples project to start new Twister2 Local cluster and run the WordCount example on it.

    $ mvn package exec:java \
        -DskipTests \
        -Dexec.mainClass=org.apache.beam.examples.WordCount \
        -Dexec.args="\
            --runner=Twister2Runner \
            --inputFile=pom.xml \
            --output=counts" \
        -Ptwister2-runner

Running WordCount on a Twister2 Deployment

The Beam examples project, when generated from an archetype, comes from a particular released Beam version (that's what the archetypeVersion property is about). Each Beam version that contains the Twister2 Runner (i.e. from 2.23.0 onwards) uses a certain version of Twister2. Because of this, when we start a stand-alone Twister2 cluster and try to run Beam examples on it we need to make sure the two are compatible. See following table for which Twister2 version is recommended for various Beam versions.

Download latest Twister2 version compatible with the Beam you are using from Twister2 Website. Twister2 currently supports several deployment options, such as standalone, Slurm, Mesos, Nomad, etc. To learn more about the Twister2 deployments and how to get them setup visit Twister2 Docs.

Issue following command in the Beam examples project to start new Twister2 job, The “twister2Home” should point to the home directory of the Twister2 standalone deployment.

Note: Currently file paths need to be absolute paths.

    $ mvn package exec:java \
        -DskipTests \
        -Dexec.mainClass=org.apache.beam.examples.WordCount \
        -Dexec.args="\
            --runner=Twister2Runner \
            --twister2Home=<PATH_TO_TWISTER2_HOME>
            --parallelism=2
            --inputFile=<PATH_TO_FILE>/pom.xml \
            --output=<PATH_TO_FILE>/counts" \
        -Ptwister2-runner

Pipeline Options for the Twister2 Runner