hadoop103kafka挂了 kafka到hdfs用flume

一、概况

在Kafka集群实际应用中，Kafka的消费者有很多种（如：应用程序、Flume、Spark Streaming、Storm等），本篇文章主要讲述如何在Kerberos环境使用Flume采集Kafka数据并写入HDFS。本文的数据流图如下：

• 内容概述
  1.Kafka集群启用Kerberos
  2.环境准备及配置Flume Agent
  3.java访问并测试
• 测试环境
  1.CM和CDH版本为5.11.2
  2.采用root用户操作
• 前置条件
  1.集群已启用Kerberos
  2.集群已安装Kafka
  3.集群已安装Flume

二、Kafka集群启用Kerberos

登录Cloudera Manager进入Kafka服务，修改如下配置Kerberos.auth.enable和security.inter.broker.protocol配置为如下截图：

保存配置并重启Kafka服务。

三、环境准备

由于Kafka集群已启用Kerberos认证，这里需要准备访问Kafka集群的环境，如Keytab、jaas.conf配置等

1.生成访问Kafka集群的keytab文件,在Kerberos所在服务上执行如下命令

[root@ip-172-31-22-86 kafkatest]# pwd
/home/ec2-user/kafkatest
[root@ip-172-31-22-86 kafkatest]# kadmin.local
Authenticating as principal hdfs/[e-mail] with password.
kadmin.local:  xst -norandkey -k fayson.keytab [e-mail]

可以看到在当前目录下生成了[e-mail]账号的keytab文件。

2.创建jaas.conf文件，文件内容如下

KafkaClient {
   com.sun.security.auth.module.Krb5LoginModule required
   useKeyTab=true
   keyTab="/keytab/fayson.keytab"
   principal="[e-mail]";
};

3.将keytab文件和jaas.conf文件拷贝至所有Flume Agent运行的节点

这里我们将上面的配置文件拷贝放在Flume Agent节点的/flume-keytab目录下

4.修改目录文件属主，确保flume用户有权限访问

[ec2-user@ip-172-31-21-45 flume-keytab]$ sudo chown -R flume. /flume-keytab/
[ec2-user@ip-172-31-21-45 flume-keytab]$ sudo chmod -R 755 /flume-keytab/

四、配置Flume Agent

1.配置Flume Agent读取Kafka数据写入HDFS

kafka.channels = c1
kafka.sources = s1
kafka.sinks = k1

kafka.sources.s1.type =org.apache.flume.source.kafka.KafkaSource
kafka.sources.s1.kafka.bootstrap.servers =ip-172-31-26-80.ap-southeast-1.compute.internal:9092,ip-172-31-21-45.ap-southeast-1.compute.internal:9092,  ip-172-31-26-102.ap-southeast-1.compute.internal:9092
kafka.sources.s1.kafka.topics = test4
kafka.sources.s1.kafka.consumer.group.id =flume-consumer
kafka.sources.s1.kafka.consumer.security.protocol= SASL_PLAINTEXT
kafka.sources.s1.kafka.consumer.sasl.mechanism= GSSAPI
kafka.sources.s1.kafka.consumer.sasl.kerberos.service.name= kafka
kafka.sources.s1.channels = c1

kafka.channels.c1.type = memory

kafka.sinks.k1.type = hdfs
kafka.sinks.k1.channel = c1
kafka.sinks.k1.hdfs.kerberosKeytab= /flume-keytab/fayson.keytab
kafka.sinks.k1.hdfs.kerberosPrincipal= [e-mail]
kafka.sinks.k1.hdfs.path =/tmp/kafka-test
kafka.sinks.k1.hdfs.filePrefix = events-
kafka.sinks.k1.hdfs.writeFormat = Text

关于HDFS Sink的更多配置可以参考：http://flume.apache.org/FlumeUserGuide.html#hdfs-sink

2.增加Flume Agent启动参数

-Djava.security.auth.login.config=/flume-keytab/jaas.conf

配置完成后保存更改并重启FlumeAgent服务。

五、Java生产消息

1.编写jaas.conf文件

KafkaClient {
  com.sun.security.auth.module.Krb5LoginModule required
  useKeyTab=true
  keyTab="/home/ec2-user/run-kafka/conf/fayson.keytab"
  principal="[e-mail]";
};

2.使用Java编写消息生产代码

package com.cloudera;

import org.apache.kafka.clients.producer.KafkaProducer;
import org.apache.kafka.clients.producer.Producer;
import org.apache.kafka.clients.producer.ProducerConfig;
import org.apache.kafka.clients.producer.ProducerRecord;

import java.io.File;
import java.util.Properties;

/**
 * package: com.cloudera
 * describe: TODO
 * creat_user: Fayson
 * email: [e-mail]
 * creat_date: 2017/12/12
 * creat_time: 下午3:35
 * 公众号：Hadoop实操
 */
public class ProducerTest {

    public static String TOPIC_NAME = "test4";

    public static String confPath = System.getProperty("user.dir") + File.separator + "conf";

    public static void main(String[] args) {
        try {
            String krb5conf = confPath + File.separator + "krb5.conf";
            String jaasconf = confPath + File.separator + "jaas.conf";

            System.setProperty("java.security.krb5.conf", krb5conf);
            System.setProperty("java.security.auth.login.config", jaasconf);
            System.setProperty("javax.security.auth.useSubjectCredsOnly", "false");
//            System.setProperty("sun.security.krb5.debug", "true"); //Kerberos Debug模式

            Properties props = new Properties();
            props.put(ProducerConfig.BOOTSTRAP_SERVERS_CONFIG, "ip-172-31-21-45.ap-southeast-1.compute.internal:9092,ip-172-31-26-102.ap-southeast-1.compute.internal:9020,ip-172-31-26-80.ap-southeast-1.compute.internal:9020");
            props.put(ProducerConfig.ACKS_CONFIG, "all");
            props.put(ProducerConfig.KEY_SERIALIZER_CLASS_CONFIG, "org.apache.kafka.common.serialization.StringSerializer");
            props.put(ProducerConfig.VALUE_SERIALIZER_CLASS_CONFIG, "org.apache.kafka.common.serialization.StringSerializer");
            props.put("security.protocol", "SASL_PLAINTEXT");
            props.put("sasl.kerberos.service.name", "kafka");

            Producer<String, String> producer = new KafkaProducer<String, String>(props);
            for (int i = 0; i < 10; i++) {
                String message = i + "\t" + "fayson" + i  + "\t" + 22+i;
                ProducerRecord record = new ProducerRecord<String, String>(TOPIC_NAME, message);
                producer.send(record);
                System.out.println(message);
            }

            producer.flush();
            producer.close();

        } catch (Exception e) {
            e.printStackTrace();
        }
    }
}

3、将工程编译打包kafka-demo-1.0-SNAPSHOT.jar

mvn clean package

4.使用mvn命令将工程依赖库导出

mvn dependency:copy-dependencies -DoutputDirectory=/Users/fayson/Desktop/lib

将导出的jar包放在run-kafka/lib目录下。

5.编写run.sh脚本，运行测试jar包

#!/bin/bash

JAVA_HOME=/usr/java/jdk1.8.0_131-cloudera

for file in `ls lib/*jar`
do
    CLASSPATH=$CLASSPATH:$file
done

export CLASSPATH

${JAVA_HOME}/bin/java com.cloudera.ProducerTest

6.conf目录文件

fayson.keytab：fayson的keytab文件

jaas.conf：java访问Kerberos环境下的配置

krb5.conf：集群的krb5配置文件

六、Kafka->Flume->HDFS流程测试

1.将第5章开发好的示例放在集群的服务器上

2.执行run.sh

[ec2-user@ip-172-31-22-86 run-kafka]$ sh run.sh

3.查看HDFS的/extwarehouse/student目录下数据

这里可以看到数据已写入HDFS指定的目录。