Java技术栈 —— Spark入门（二）之实时WordCount

参考文章：

实验环境：
假设你的用户为root，以下软件安装路径为/opt

实验结构图

一、kafka

1.1 创建topic

# 创建input
bin/kafka-topics.sh --bootstrap-server localhost:9092 --create --topic test.wordcount.input --partitions 1 --replication-factor 1
# 创建output
bin/kafka-topics.sh --bootstrap-server localhost:9092 --create --topic test.wordcount.output --partitions 1 --replication-factor 1

1.2 准备input与查看output

# 打开两个terminal终端
# 准备键盘输入作为prodcuer
bin/kafka-console-producer.sh --broker-list localhost:9092 --topic test.wordcount.input
# 在屏幕上查看输出
bin/kafka-console-consumer.sh --bootstrap-server localhost:9092 --topic test.wordcount.output

二、spark

2.1 spark下的程序文件

# coding=utf-8
# /opt/spark-3.5.2-bin-hadoop3/jobs/pyjobs/kafka-wordcount.py
from __future__ import print_function
from pyspark.sql import SparkSession
from pyspark.sql.functions import explode
from pyspark.sql.functions import split
from pyspark.sql import functions as F

bootstrapServers = "localhost:9092"

spark = SparkSession\
    .builder\
    .appName("StructuredKafkaWordCount")\
    .getOrCreate()

# 基于来自kafka的数据流，创建dataframe
lines = spark\
    .readStream\
    .format("kafka")\
    .option("kafka.bootstrap.servers", bootstrapServers)\
    .option("subscribe", "test.wordcount.input")\
    .option("failOnDataLoss", False)\
    .option("group.id", "wordcount-group3")\
    .load()\
    .selectExpr("CAST(value AS STRING)")

# 将单行数据拆分，转成多行数据
words = lines.select(
    explode(split(lines.value, ' ')).alias('word')
)

# 对单词进行分组，并计算总数
wordCounts = words.groupBy('word').count()

# 将两列数据合并成单列数据
wordCounts = wordCounts.select(F.concat(F.col("word"), F.lit("|"), F.col("count").cast("string")).alias("value"))

# 测试时，可以不将结果写入kafka，直接输出到控制台
# query = wordCounts \
#     .writeStream \
#     .outputMode("complete") \
#     .format("console") \
#     .start()

# 将结果输出到 test.wordcount.output
query = wordCounts \
    .writeStream \
    .format('kafka') \
    .outputMode('update') \
    .option("kafka.bootstrap.servers", bootstrapServers) \
    .option('checkpointLocation', '/spark/job-checkpoint') \
    .option("topic", "test.wordcount.output") \
    .start()

query.awaitTermination()

2.2 用spark-submit提交作业

# 提交Spark作业，这个过程需要保证网络畅通，会将一些依赖下载到/root/.ivy2/jars目录下
$SPARK_HOME/bin/spark-submit \
--packages org.apache.spark:spark-sql-kafka-0-10_2.12:3.5.2,\
org.apache.kafka:kafka-clients:3.5.2 \
/opt/spark-3.5.2-bin-hadoop3/jobs/pyjobs/kafka-wordcount.py