背景说明

公司使用华为提供的大数据集群管理公司数据
集群的组件使用了kerberos认证，现在想使用程序和setreamsets访问集群kafka需要通过认证
集群kafka加入了华为自己的安全相关的东西，要访问kafka，必须要用华为提供的客户端包!
华为为外部访问的组件和程序提供认证所需的东西: keytab,krb5.conf,principal

注意：在集群外访问华为集群，除了打通网络，还要开启认证相关的端口
查看krb5.conf,找到每个认证域下的 kdc = xxx，这个xxx后面的端口是UDP端口，需要将UDP端口开启！

java代码认证

华为官方提供了代码认证的实例，可以按照这个来，示例代码地址: https://github.com/huaweicloud/huaweicloud-mrs-example
也可以根据通用的认证模式来写测试代码，只要kafka客户端使用华为提供的就行
示例代码:

public class App {  
  
    private static final Logger logger = LoggerFactory.getLogger(App.class);  
  
    public static void setBootstrapServer(Properties props) {  
        props.put("bootstrap.servers", "k1:9092,k2:9092,k3:9092");        
    }  
  
    public static void main(String[] args) {  
        // 认证相关配置文件指定  
        setSystemProperty();  
  
        Properties props = new Properties();  
        setBootstrapServer(props); // kafka地址配置  
        setCommonKafkaProps(props); // kafka通用配置  
        securityAuth(props); // 认证相关配置  
  
        System.out.println("begin send message to kafka!");  
        Producer<String, String> producer;  
        try {  
           producer  = new KafkaProducer<>(props);  
        } catch (Exception e) {  
            logger.error("创建producer失败:", e);  
            for (StackTraceElement element : e.getStackTrace()) {  
                System.out.println(element);  
            }  
            Throwable cause = e.getCause();  
            while (cause != null) {  
                System.out.println("Caused by: " + cause);  
                for (StackTraceElement element : cause.getStackTrace()) {  
                    System.out.println(element);  
                }  
                cause = cause.getCause();  
            }  
            return;  
        }  
        System.out.println("创建生产者成功!");  
  
        long timeBegin = System.currentTimeMillis();  
        String topic = "test_tpc";  
        for (int i = 0; i < 10; i++) {  
            String msg = "send one message: " + i;  
            logger.info("准备发送消息: {}", msg);  
            producer.send(new ProducerRecord<String, String>(topic, msg), new SendCallBack());  
            logger.info("send to kafka finish, message is: {}", msg);  
        }  
        logger.info("send message success, cost time: {}", (System.currentTimeMillis() - timeBegin) / 1000);  
        producer.flush();  
        producer.close();  
  
        logger.info("kafka send msg to topic finished!");  
    }  
  
    public static void setSystemProperty() {  
        System.setProperty("zookeeper.server.principal", "zookeeper/hadoop.hadoop.com");  
        System.setProperty("sun.security.krb5.debug", "true");  
        System.setProperty("java.security.auth.login.config", "/xxx/jaas.conf"); 
        System.setProperty("java.security.krb5.conf", "/xxx/krb5.conf");  
    }  
  
    public static void setCommonKafkaProps(Properties props) {  
        props.put("acks", "all");  
        props.put("retries", 0);  
        props.put("batch.size", 16384);  
        props.put("linger.ms", 1);  
        props.put("buffer.memory", 33554432);  
        props.put("key.serializer", "org.apache.kafka.common.serialization.StringSerializer");  
        props.put("value.serializer", "org.apache.kafka.common.serialization.StringSerializer");  
    }  
  
    public static void securityAuth(Properties props) {  
        // kerberos 安全认证  
        props.put("security.protocol", "SASL_PLAINTEXT");  
        props.put("sasl.mechanism", "GSSAPI");  
        props.put("sasl.kerberos.service.name", "kafka");  
//        props.put("kerberos.domain.name", "hadoop.hadoop.com");  
    }  
}

pom.xml 引用华为相关包，这个可以抄示例代码，具体依赖就不列出来了

   <properties>  
        <zookeeper.version>3.6.3-h0.cbu.mrs.320.r33</zookeeper.version>  
        <kafka.version>2.4.0-h0.cbu.mrs.320.r33</kafka.version>  
        <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>  
    </properties>  

代码中使用的jaas.conf如下:

KafkaClient {
   com.sun.security.auth.module.Krb5LoginModule required
   useKeyTab=true
   principal="[email protected]"
   keyTab="/xxx/your.keytab"
   storeKey=true
   useTicketCache=false
   debug=true;
};

Client {
   com.sun.security.auth.module.Krb5LoginModule required
   useKeyTab=true
   principal="[email protected]"
   keyTab="/xxx/your.keytab"
   storeKey=true
   useTicketCache=false
   debug=true;
};

streamsets认证

修改streamsets配置：
找到streamsets配置文件 streamsets/etc/sdc.properties，修改下面三个

# Runs the Data Collector within a Kerberos session which is propagated to all stages.
# This is useful for stages that require Kerberos authentication with the services they interact with
# kerberos.client.enabled=false
kerberos.client.enabled=true

# The Kerberos principal to use for the Kerberos session.
# It should be a service principal. If the hostname part of the service principal is '_HOST' or '0.0.0.0',
# the hostname will be replaced with the actual complete hostname of Data Collector as advertised by the
# unix command 'hostname -f'
# kerberos.client.principal=sdc/[email protected]
[email protected]


# The location of the keytab file for the specified principal. If the path is relative, the keytab file will be
# looked for in the Data Collector configuration directory
# kerberos.client.keytab=sdc.keytab
kerberos.client.keytab=/xxx/your.keytab

启动时添加启动参数,指定jaas配置，和上面代码中配置一样

SDC_JAVA_OPTS="-Djava.security.auth.login.config=/xxx/jaas.conf" ./streamsets dc

访问kafka时，因使用的是华为的包，需要在kafka配置中上传华为的kafka相关包
并添加认证相关的配置

security.protocol = SASL_PLAINTEXT
sasl.mechanism = GSSAPI
sasl.kerberos.service.name = kafka