Instance

Scenario

Assume that the Kafka component receives one word record every 1 second.

The developed Spark application needs to achieve the following function:

Calculate the sum of records for each word in real time.

log1.txt example file:

LiuYang
YuanJing
GuoYijun
CaiXuyu
Liyuan
FangBo
LiuYang
YuanJing
GuoYijun
CaiXuyu
FangBo

Data Planning

Spark Streaming sample project data is stored in the Kafka component. A user with Kafka permission sends data to Kafka component.

Ensure that the cluster, including HDFS, Yarn, Spark, and Kafka is successfully installed.
Create the input_data1.txt file in the local and copy the content of the log1.txt file to the input_data1.txt file.
On the client installation node, create the /home/data directory and upload the input_data1.txt file to the /home/data directory.
Create a topic.
{zkQuorum} indicates ZooKeeper cluster information. The format is IP:port.

$KAFKA_HOME/bin/kafka-topics.sh --create --zookeeper {zkQuorum}/kafka --replication-factor 1 --partitions 3 --topic {Topic}
Start the Producer of Kafka and send data to Kafka.
java -cp {ClassPath} com.huawei.bigdata.spark.examples.StreamingExampleProducer {BrokerList} {Topic}

In this command, ClassPath must include the absolute path of the Kafka JAR package of the Spark client, for example: /opt/client/Spark2x/spark/jars/*:/opt/client/Spark2x/spark/jars/streamingClient010/*

Development Approach

Receive data from Kafka and generate DStream.
Collect the statistics of word records by category.
Calculate and print the result.

Packaging the Project

Use the Maven tool provided by IDEA to pack the project and generate a JAR file. For details, see Compiling and Running the Application.
Upload the JAR package to any directory (for example, /opt) on the server where the Spark client is located.

For the dependency package whose version number contains "hw-ei", download from the Huawei open-source image site.
For the dependency package whose version number does not contain "hw-ei", obtain them from the Maven central repository.

Running Tasks

When running the sample program, you need to specify <checkpointDir>, <brokers>, <topic>, and <batchTime>. <checkPointDir> indicates the path for storing the program result backup in HDFS. <brokers> indicates the Kafka address for obtaining metadata. <topic> indicates the topic name read from Kafka. <batchTime> indicates the interval for Streaming processing in batches.

The location of Spark Streaming Kafka dependency package on the client is different from the location of other dependency packages. For example, the path to the Spark Streaming Kafka dependency package is $SPARK_HOME/jars/streamingClient010, whereas the path to other dependency packages is $SPARK_HOME/jars. Therefore, when running an application, you need to add a configuration item to the spark-submit command to specify the path of the dependency package of Spark Streaming Kafka, for example, --jars $(files=($SPARK_HOME/jars/streamingClient010/*.jar); IFS=,; echo "${files[*]}")

Go to the Spark client directory and run the following commands to invoke the bin/spark-submit script to run the code (The class name and file name must be the same as those in the actual code. The following is only an example):

Example code (Spark Streaming read Kafka 0-10 Write To Print)
bin/spark-submit --master yarn --deploy-mode client --jars$(files=($SPARK_HOME/jars/streamingClient010/*.jar); IFS=,; echo "${files[*]}") --class com.huawei.bigdata.spark.examples.KafkaWordCount /opt/SparkStreamingKafka010JavaExample-1.0.jar <checkpointDir> <brokers> <topic> <batchTime>
Spark Streaming Write To Kafka 0-10 example code:
bin/spark-submit --master yarn --deploy-mode client --jars $(files=($SPARK_HOME/jars/streamingClient010/*.jar); IFS=,; echo "${files[*]}") --class com.huawei.bigdata.spark.examples.JavaDstreamKafkaWriter /opt/SparkStreamingKafka010JavaExample-1.0.jar <groupId> <brokers> <topics>