Apache samza

Clone this repo:

Branches

  1. 09a2968 Update scripts and docs based on Beam 2.11.0 release #2 by xinyuiscool · 7 months ago master
  2. 0243d04 Update scripts and docs based on Beam 2.11.0 release by xinyuiscool · 7 months ago
  3. 5d02a5b Add README by xiliu · 9 months ago
  4. a4addbd Add yarn examples by xinyuiscool · 9 months ago
  5. e5f7072 Add a Kafka word count example by xiliu · 10 months ago

Apache Beam Examples Using SamzaRunner

The examples in this repository serve to demonstrate running Beam pipelines with SamzaRunner locally, in Yarn cluster, or in standalone cluster with Zookeeper. More complex pipelines can be built from here and run in similar manner.

Example Pipelines

The following examples are included:

  1. WordCount reads a file as input (bounded data source), and computes word frequencies.

  2. KafkaWordCount does the same word-count computation but reading from a Kafka stream (unbounded data source). It uses a fixed 10-sec window to aggregate the counts.

Run the Examples

Each example can be run locally, in Yarn cluster or in standalone cluster. Here we use KafkaWordCount as an example.

Set Up

  1. Download and install JDK version 8. Verify that the JAVA_HOME environment variable is set and points to your JDK installation.

  2. Download and install Apache Maven by following Maven’s installation guide for your specific operating system.

Check out the samza-beam-examples repo:

$ git clone https://github.com/apache/samza-beam-examples.git
$ cd samza-beam-examples

A script named “grid” is included in this project which allows you to easily download and install Zookeeper, Kafka, and Yarn. You can run the following to bring them all up running in your local machine:

$ scripts/grid bootstrap

All the downloaded package files will be put under deploy folder. Once the grid command completes, you can verify that Yarn is up and running by going to http://localhost:8088. You can also choose to bring them up separately, e.g.:

$ scripts/grid install zookeeper
$ scripts/grid start zookeeper

Now let's create a Kafka topic named “input-text” for this example:

$ ./deploy/kafka/bin/kafka-topics.sh  --zookeeper localhost:2181 --create --topic input-text --partitions 10 --replication-factor 1

Run Locally

You can run directly within the project using maven:

$ mvn compile exec:java -Dexec.mainClass=org.apache.beam.examples.KafkaWordCount \
    -Dexec.args="--runner=SamzaRunner" -P samza-runner

Packaging Your Application

To execute the example in either Yarn or standalone, you need to package it first. After packaging, we deploy and explode the tgz in the deploy folder:

 $ mkdir -p deploy/examples
 $ mvn package && tar -xvf target/samza-beam-examples-0.1-dist.tar.gz -C deploy/examples/

Run in Standalone Cluster with Zookeeper

You can use the run-beam-standalone.sh script included in this repo to run an example in standalone mode. The config file is provided as config/standalone.properties. Note by default we create one single split for the whole input (--maxSourceParallelism=1). To set each Kafka partition in a split, we can set a large “maxSourceParallelism” value which is the upper bound of the number of splits.

$ deploy/examples/bin/run-beam-standalone.sh org.apache.beam.examples.KafkaWordCount \
    --configFilePath=$PWD/deploy/examples/config/standalone.properties --maxSourceParallelism=1024

Run Yarn Cluster

Similar to running standalone, we can use the run-beam-yarn.sh to run the examples in Yarn cluster. The config file is provided as config/yarn.properties. To run the KafkaWordCount example in yarn:

$ deploy/examples/bin/run-beam-yarn.sh org.apache.beam.examples.KafkaWordCount \
    --configFilePath=$PWD/deploy/examples/config/yarn.properties --maxSourceParallelism=1024

Validate the Pipeline Results

Now the pipeline is deployed to either locally, standalone or Yarn. Let's check out the results. First we start a kakfa consumer to listen to the output:

$ ./deploy/kafka/bin/kafka-console-consumer.sh --bootstrap-server localhost:9092 --topic word-count --property print.key=true

Then let's publish a few lines to the input Kafka topic:

$ ./deploy/kafka/bin/kafka-console-producer.sh --topic input-text --broker-list localhost:9092
Nory was a Catholic because her mother was a Catholic, and Nory’s mother was a Catholic because her father was a Catholic, and her father was a Catholic because his mother was a Catholic, or had been.

You should see the word count shows up in the consumer console in about 10 secs:

a       6
br      1
mother  3
was     6
Catholic        6
his     1
Nory    2
s       1
father  2
had     1
been    1
and     2
her     3
or      1
because 3

Beyond Examples

Feel free to build more complex pipelines based on the examples above, and reach out to us:

More Information