Flink Connector Kafka
- 1. Kafka
- 1.1. [Kafka官网](http://kafka.apache.org/)
- 1.2. Kafka 简述
- 1.3. Kafka特性
- 1.4. kafka的应用场景
- 1.5. kafka-manager的部署
- 1.6. `使用Kafka Connect导入/导出数据`
- 1.7. [Kafka日志存储原理]()
- 2. Kafka与Flink的融合
- 2.1. kafka连接flink流计算,实现flink消费kafka的数据
- 2.2. flink 读取kafka并且自定义水印再将数据写入kafka中
- 3. Airbnb 是如何通过 balanced Kafka reader 来扩展 Spark streaming 实时流处理能力的
- 4. 寄语:海阔凭鱼跃,天高任鸟飞
1. Kafka
1.1. Kafka官网
1.2. Kafka 简述
- Kafka 是一个分布式消息系统:具有生产者、消费者的功能。它提供了类似于JMS 的特性,但是在设计实现上完全不同,此外它并不是JMS 规范的实现。
1.3. Kafka特性
-
消息持久化
:基于文件系统来存储和缓存消息 高吞吐量
-
多客户端支持
:核心模块用Scala
语言开发,Kafka 提供了多种开发语言的接入,如Java 、Scala、C 、C++、Python 、Go 、Erlang 、Ruby 、Node. 等 安全机制
通过SSL 和SASL(Kerberos), SASL/PLA时验证机制支持生产者、消费者与broker连接时的身份认证;
支持代理与ZooKeeper 连接身份验证
通信时数据加密
客户端读、写权限认证
Kafka 支持与外部其他认证授权服务的集成
数据备份
轻量级
消息压缩
1.4. kafka的应用场景
Kafka作为消息传递系统
Kafka 作为存储系统
Kafka用做流处理
消息,存储,流处理结合起来使用
1.5. kafka-manager的部署
Kafka Manager 由 yahoo 公司开发,该工具可以方便查看集群 主题分布情况,同时支持对 多个集群的管理、分区平衡以及创建主题等操作。
- Centos7安装kafka-manager
启动脚本
bin/cmak -Dconfig.file=conf/application.conf -java-home /usr/lib/jdk-11.0.6 -Dhttp.port=9008 &
- 界面效果
注意
1.6. 使用Kafka Connect导入/导出数据
- 替代Flume——Kafka Connect
集群模式
注意: 在集群模式下,配置并不会在命令行传进去,而是需要REST API来创建,修改和销毁连接器。
- 通过一个示例了解kafka connect连接器
- kafka connect简介以及部署
1.7. Kafka日志存储原理
Kafka的Message存储采用了分区(partition)
,分段(LogSegment)
和稀疏索引
这几个手段来达到了高效性
- 查看分区.index文件
bin/kafka-run-class.sh kafka.tools.DumpLogSegments --files kafka-logs/t2-2/00000000000000000000.index
- 查看log文件
/bin/kafka-run-class.sh kafka.tools.DumpLogSegments --files t1-1/00000000000000000000.log --print-data-log
- 查看TimeIndex文件
bin/kafka-run-class.sh kafka.tools.DumpLogSegments --files t1-2/00000000000000000000.timeindex --verify-index-only
- 引入时间戳的作用
2. Kafka与Flink的融合
Flink 提供了专门的 Kafka 连接器,向 Kafka topic 中读取或者写入数据。Flink Kafka Consumer 集成了 Flink 的 Checkpoint 机制,可提供 exactly-once 的处理语义。为此,Flink 并不完全依赖于跟踪 Kafka 消费组的偏移量,而是在内部跟踪和检查偏移量。
2.1. kafka连接flink流计算,实现flink消费kafka的数据
- 创建flink项目
sbt new tillrohrmann/flink-project.g8
- 配置sbt
ThisBuild / resolvers ++= Seq(
"Apache Development Snapshot Repository" at "https://repository.apache.org/content/repositories/snapshots/",
Resolver.mavenLocal
)
name := "FlinkKafkaProject"
version := "1.0"
organization := "com.xiaofan"
ThisBuild / scalaVersion := "2.12.6"
val flinkVersion = "1.10.0"
val kafkaVersion = "2.2.0"
val flinkDependencies = Seq(
"org.apache.flink" %% "flink-scala" % flinkVersion % "provided",
"org.apache.kafka" %% "kafka" % kafkaVersion % "provided",
"org.apache.flink" %% "flink-connector-kafka" % flinkVersion,
"org.apache.flink" %% "flink-streaming-scala" % flinkVersion % "provided")
lazy val root = (project in file(".")).
settings(
libraryDependencies ++= flinkDependencies
)
assembly / mainClass := Some("com.xiaofan.Job")
// make run command include the provided dependencies
Compile / run := Defaults.runTask(Compile / fullClasspath,
Compile / run / mainClass,
Compile / run / runner
).evaluated
// stays inside the sbt console when we press "ctrl-c" while a Flink programme executes with "run" or "runMain"
Compile / run / fork := true
Global / cancelable := true
// exclude Scala library from assembly
assembly / assemblyOption := (assembly / assemblyOption).value.copy(includeScala = false)
- 源代码
package com.xiaofan
import java.util.Properties
import org.apache.flink.api.common.serialization.SimpleStringSchema
import org.apache.flink.streaming.api.scala.{StreamExecutionEnvironment, _}
import org.apache.flink.streaming.api.{CheckpointingMode, TimeCharacteristic}
import org.apache.flink.streaming.connectors.kafka.FlinkKafkaConsumer
/**
* 用flink消费kafka
*
* @author xiaofan
*/
object ReadingFromKafka {
val ZOOKEEPER_HOST = "192.168.1.23:2181,192.168.1.24:2181,192.168.1.25:2181"
val KAFKA_BROKER = "192.168.1.23:9091,192.168.1.24:9091,192.168.1.25:9091"
val TRANSACTION_GROUP = "com.xiaofan.flink"
def main(args: Array[String]): Unit = {
val env: StreamExecutionEnvironment = StreamExecutionEnvironment.getExecutionEnvironment
env.setStreamTimeCharacteristic(TimeCharacteristic.EventTime)
env.enableCheckpointing(1000)
env.getCheckpointConfig.setCheckpointingMode(CheckpointingMode.EXACTLY_ONCE)
// configure kafka consumer
val kafkaProps = new Properties()
kafkaProps.setProperty("zookeeper.connect", ZOOKEEPER_HOST)
kafkaProps.setProperty("bootstrap.servers", KAFKA_BROKER)
kafkaProps.setProperty("group.id", TRANSACTION_GROUP)
val transaction: DataStream[String] = env.addSource(new FlinkKafkaConsumer[String]("xiaofan01", new SimpleStringSchema(), kafkaProps))
transaction.print
env.execute()
}
}
- 启动kafka集群,运行结果
2.2. flink 读取kafka并且自定义水印再将数据写入kafka中
- 需求说明(自定义窗口,每分钟的词频统计)
- 从kafka中读取数据(topic:t1)
- kafka中有event time时间值,通过该时间戳来进行时间划分,窗口长度为10秒,窗口步长为5秒
- 由于生产中可能会因为网络或者其他原因导致数据延时,比如 00:00:10 时间的数据可能 00:00:12 才会传入kafka中,所以在flink的处理中应该设置延时等待处理,这里设置的2秒,可以自行修改。
- 结果数据写入kafka中(topic:t2)(数据格式
time:时间 count:每分钟的处理条数
)
- 准备环境flink1.10.0 + kafka2.2.0
- 创建topic
bin/kafka-topics.sh --create --bootstrap-server 192.168.1.25:9091 --replication-factor 2 --partitions 3 --topic t1
bin/kafka-topics.sh --create --bootstrap-server 192.168.1.25:9091 --replication-factor 2 --partitions 3 --topic t2
- 向t1中生产数据
package com.xiaofan
import java.util.Properties
import org.apache.kafka.clients.producer.{KafkaProducer, ProducerRecord}
object ProduceData {
def main(args: Array[String]): Unit = {
val props = new Properties()
props.put("bootstrap.servers", "192.168.1.25:9091")
props.put("acks", "1")
props.put("retries", "3")
props.put("batch.size", "16384") // 16K
props.put("linger.ms", "1")
props.put("buffer.memory", "33554432") // 32M
props.put("key.serializer", "org.apache.kafka.common.serialization.StringSerializer")
props.put("value.serializer", "org.apache.kafka.common.serialization.StringSerializer")
val producer = new KafkaProducer[String, String](props)
var i = 0
while (true) {
i += 1
// 模拟标记事件时间
val record = new ProducerRecord[String, String]("t1", i + "," + System.currentTimeMillis())
// 只管发送消息,不管是否发送成功
producer.send(record)
Thread.sleep(300)
}
}
}
- 消费t1数据,处理后再次传入kafka t2
package com.xiaofan
import java.text.SimpleDateFormat
import java.util.{Date, Properties}
import org.apache.flink.api.common.serialization.SimpleStringSchema
import org.apache.flink.streaming.api.TimeCharacteristic
import org.apache.flink.streaming.api.functions.AssignerWithPeriodicWatermarks
import org.apache.flink.streaming.api.scala.{StreamExecutionEnvironment, _}
import org.apache.flink.streaming.api.watermark.Watermark
import org.apache.flink.streaming.api.windowing.time.Time
import org.apache.flink.streaming.connectors.kafka.internals.KeyedSerializationSchemaWrapper
import org.apache.flink.streaming.connectors.kafka.{FlinkKafkaConsumer, FlinkKafkaProducer}
/**
* Watermark 案例
* 根据自定义水印定义时间,计算每秒的消息数并且写入 kafka中
*/
object StreamingWindowWatermarkScala {
def main(args: Array[String]): Unit = {
val env = StreamExecutionEnvironment.getExecutionEnvironment
env.setStreamTimeCharacteristic(TimeCharacteristic.EventTime)
env.setParallelism(1)
val topic = "t1"
val prop = new Properties()
prop.setProperty("bootstrap.servers","192.168.1.25:9091")
prop.setProperty("group.id","con1")
val myConsumer = new FlinkKafkaConsumer[String](topic,new SimpleStringSchema(),prop)
// 添加源
val text = env.addSource(myConsumer)
val inputMap = text.map(line=>{
val arr = line.split(",")
(arr(0),arr(1).trim.toLong)
})
// 添加水印
val waterMarkStream = inputMap.assignTimestampsAndWatermarks(new AssignerWithPeriodicWatermarks[(String, Long)] {
var currentMaxTimestamp = 0L
var maxOutOfOrderness = 3000L// 最大允许的乱序时间是10s
val sdf = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss.SSS");
override def getCurrentWatermark = new Watermark(currentMaxTimestamp - maxOutOfOrderness)
override def extractTimestamp(element: (String, Long), previousElementTimestamp: Long) = {
val timestamp = element._2
currentMaxTimestamp = Math.max(timestamp, currentMaxTimestamp)
val id = Thread.currentThread().getId
println("currentThreadId:"+id+",key:"+element._1+",eventtime:["+element._2+"|"+sdf.format(element._2)+"],currentMaxTimestamp:["+currentMaxTimestamp+"|"+ sdf.format(currentMaxTimestamp)+"],watermark:["+getCurrentWatermark().getTimestamp+"|"+sdf.format(getCurrentWatermark().getTimestamp)+"]")
timestamp
}
})
val window = waterMarkStream.map(x=>(x._2,1)).timeWindowAll(Time.seconds(1),Time.seconds(1)).sum(1).map(x=>"time:"+tranTimeToString(x._1.toString)+" count:"+x._2)
// .window(TumblingEventTimeWindows.of(Time.seconds(3))) //按照消息的EventTime分配窗口,和调用TimeWindow效果一样
val topic2 = "t2"
val props = new Properties()
props.setProperty("bootstrap.servers","192.168.1.25:9091")
//使用支持仅一次语义的形式
val myProducer = new FlinkKafkaProducer[String](topic2,new KeyedSerializationSchemaWrapper[String](new SimpleStringSchema()), props, FlinkKafkaProducer.Semantic.EXACTLY_ONCE)
window.addSink(myProducer)
env.execute("StreamingWindowWatermarkScala")
}
def tranTimeToString(timestamp:String) :String={
val fm = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss")
val time = fm.format(new Date(timestamp.toLong))
time
}
}
- 运行效果
3. Airbnb 是如何通过 balanced Kafka reader 来扩展 Spark streaming 实时流处理能力的
4. 寄语:海阔凭鱼跃,天高任鸟飞