本篇文章大概6457字,阅读时间大约17分钟
Flume作为一个数据接入组件,广泛应用于Hadoop生态中。在业务时间混乱的情况下,按照机器数据在HDFS上分区会降低ETL的效率。采用Flume自定义拦截器可以实现按照事件时间Sink到HDFS目录,以应对数据的事件时间混乱问题
1
文档编写目的
- Flume自定义拦截器的开发和测试,应对日志事件时间混乱问题
集群环境
- CDH5.16.2
2
组件介绍
Flume是一个分布式、高可靠、高可用的海量日志采集、聚合、传输系统
- Agent是一个JVM进程,控制Event从source到sink。
- Source数据源,负责数据接收
- Channel位于Source和Sink之间的buffer。Channel是线程安全的,可以同步处理多个source的写操作和多个sink的读操作
- Memory Channel基于内存,效率高,但在agent挂掉,重启等可能会有数据丢失
- File Channel基于磁盘,效率较低,不会丢数据
- Sink不断轮询Channel的事件且批量拉取,并将这些Event写入外部系统。Sink具有事务,在从Channel批量删除数据之前,每个Sink用Channel启动一个事务。批量事件一旦成功写出,sink就会进行事务提交。事务提交后,Channel从buffer中移除这批Event
- Event是Flume定义的一个数据流传输的最小单位
Flume拦截器
- Flume支持使用拦截器在运行时对event进行修改或丢弃
- Flume支持链式的拦截器执行方式,在配置文件里面配置多个拦截器,拦截器的执行顺序取决于它们配置的顺序,Event按照顺序经过每一个拦截器
3
Flume自定义拦截器实战
业务场景
在物联网的场景中,存在网络信号不佳,这时设备不会把数据传输到云平台上,而是放置在本地存储中,等待下一个开机,网络信号良好的情况下,将数据上传,造成了事件时间和平台接收时间存在跨天的情况,甚至由于设备本地时钟混乱,获取不到正确的事件时间,产生无效数据。
设备的数据上传后会进入kafka中,采用Flume拉取kafka的数据sink到HDFS接入Hive外部表进行离线分析,这里就需要使用Flume自定义拦截器按照事件时间将kafka中的数据sink到按天分区的不同的HDFS目录
实战
这里使用样例数据代替真实数据,样例数据如下:
代码语言:javascript复制2020-08-20 11:56:02.557 [main] INFO com.AppStart - {"app_active":{"name":"app_active","json":{"entry":"1","action":"1","error_code":"0"},"time":1595312507640},"attr":{"area":"石嘴山","uid":"2F10092A99995","app_v":"1.1.4","event_type":"common","device_id":"1FB872-9A10099995","os_type":"0.87","channel":"XO","language":"chinese","brand":"Huawei-0"}}
2020-08-20 11:56:02.557 [main] INFO com.AppStart - {"app_active":{"name":"app_active","json":{"entry":"1","action":"0","error_code":"0"},"time":1595312539940},"attr":{"area":"九江","uid":"2F10092A99996","app_v":"1.1.5","event_type":"common","device_id":"1FB872-9A10099996","os_type":"9.0","channel":"PU","language":"chinese","brand":"xiaomi-9"}}
自定义Flume拦截器主要就是需要实现flume的Interceptor接口,核心方法是重写intercept方法
代码语言:javascript复制public interface Interceptor {
/**
* Any initialization / startup needed by the Interceptor.
*/
public void initialize();
/**
* Interception of a single {@link Event}.
* @param event Event to be intercepted
* @return Original or modified event, or {@code null} if the Event
* is to be dropped (i.e. filtered out).
*/
public Event intercept(Event event);
/**
* Interception of a batch of {@linkplain Event events}.
* @param events Input list of events
* @return Output list of events. The size of output list MUST NOT BE GREATER
* than the size of the input list (i.e. transformation and removal ONLY).
* Also, this method MUST NOT return {@code null}. If all events are dropped,
* then an empty List is returned.
*/
public List<Event> intercept(List<Event> events);
/**
* Perform any closing / shutdown needed by the Interceptor.
*/
public void close();
/** Builder implementations MUST have a no-arg constructor */
public interface Builder extends Configurable {
public Interceptor build();
}
}
根据事件时间分区的原理就是,将设备中的事件时间解析出来,作为一个属性put到event的header中,然后在Flume的HDFS Sink配置中指定header中put的属性,代码实现如下:
代码语言:javascript复制/**
* 物联网的部分数据会保存在边缘设备上,直到下次开机进行上传,因此在用flume进行数据搜集的时候会存在补发的问题
* 落分区应该按照事件时间而不是flume主机的时间
* 事件时间拦截器则是为了应对以上场景
* @author Eights
*/
public class EventTimeInterceptor implements Interceptor {
private static FastDateFormat dateFormat = FastDateFormat.getInstance("yyyy-MM-dd");
@Override
public void initialize() {
}
@Override
public Event intercept(Event event) {
//获取header
Map<String, String> headers = event.getHeaders();
//获取body
String eventBody = new String(event.getBody(), StandardCharsets.UTF_8);
String[] bodyArr = eventBody.split("\s ");
try {
String jsonStr = bodyArr[6];
//数据为空,返回null
if (Strings.isNullOrEmpty(jsonStr)) {
return null;
}
long ts = Long.parseLong(JSON.parseObject(jsonStr).getJSONObject("app_active").getString("time"));
//打上事件日期
String eventDate = dateFormat.format(ts);
//header中添加event date
headers.put("eventDate", eventDate);
event.setHeaders(headers);
} catch (Exception e) {
//脏数据,需要sink到一个目录进行核查
headers.put("eventDate", "unknow");
event.setHeaders(headers);
}
return event;
}
@Override
public List<Event> intercept(List<Event> list) {
return list.stream().map(this::intercept)
.filter(Objects::nonNull)
.collect(Collectors.toList());
}
@Override
public void close() {
}
public static class Builder implements Interceptor.Builder {
@Override
public Interceptor build() {
return new EventTimeInterceptor();
}
@Override
public void configure(Context context) {
}
}
}
代码语言:javascript复制# pom文件
<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
<modelVersion>4.0.0</modelVersion>
<groupId>com.eights</groupId>
<artifactId>flume-ng-interceptors</artifactId>
<version>1.0-SNAPSHOT</version>
<properties>
<project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
<compiler.version>1.8</compiler.version>
<flume.version>1.9.0</flume.version>
<fastjson.version>1.2.73</fastjson.version>
</properties>
<dependencies>
<dependency>
<groupId>org.apache.flume</groupId>
<artifactId>flume-ng-core</artifactId>
<version>${flume.version}</version>
</dependency>
<dependency>
<groupId>com.alibaba</groupId>
<artifactId>fastjson</artifactId>
<version>${fastjson.version}</version>
</dependency>
</dependencies>
<build>
<plugins>
<plugin>
<artifactId>maven-compiler-plugin</artifactId>
<version>2.3.2</version>
<configuration>
<source>${compiler.version}</source>
<target>${compiler.version}</target>
</configuration>
</plugin>
<plugin>
<artifactId>maven-assembly-plugin</artifactId>
<configuration>
<descriptorRefs>
<descriptorRef>jar-with-dependencies</descriptorRef>
</descriptorRefs>
</configuration>
<executions>
<execution>
<id>make-assembly</id>
<phase>package</phase>
<goals>
<goal>single</goal>
</goals>
</execution>
</executions>
</plugin>
</plugins>
</build>
</project>
- 代码开发完成后,打包放在flume的lib目录下
- CDH集群放在/opt/cloudera/parcels/CDH/lib/flume-ng/lib,注意每个agent节点都需要配置
4
功能测试
- 将机器上的日志,通过flume sink到hdfs目录上,观察是否根据事件时间生成目录,Flume配置如下
a1.sources = r1
a1.sinks = k1
a1.channels = c1
# source
a1.sources.r1.type = TAILDIR
a1.sources.r1.positionFile =/u01/sample_data/conf/startlog_position.json
a1.sources.r1.filegroups = f1
a1.sources.r1.filegroups.f1 = /u01/sample_data/middlelog/.*log
a1.sources.r1.interceptors = i1
a1.sources.r1.interceptors.i1.type = com.eights.EventTimeInterceptor$Builder
# memorychannel
a1.channels.c1.type = memory
a1.channels.c1.capacity = 100000
a1.channels.c1.transactionCapacity = 2000
# sink
a1.sinks.k1.type = hdfs
a1.sinks.k1.hdfs.path =/ext-data/start-log/dt=%{eventDate}/
a1.sinks.k1.hdfs.filePrefix = startlog
a1.sinks.k1.hdfs.rollSize = 33554432
a1.sinks.k1.hdfs.rollCount = 0
a1.sinks.k1.hdfs.rollInterval = 0
a1.sinks.k1.hdfs.idleTimeout = 0
a1.sinks.k1.hdfs.minBlockReplicas = 1
a1.sinks.k1.hdfs.batchSize = 1000
a1.sources.r1.channels = c1
a1.sinks.k1.channel = c1
- 启动flume agent,发现hdfs sink目录按照事件时间正确创建
- 检查HDFS目录,flume自定义拦截器按照事件时间接入HDFS完成
5
总结
在未使用Flume拦截器的时候,会在数仓层面对昨天入库的数据,先按照事件时间进行重分区在做ETL,采用自定义拦截器的方式,可以直接将事件时间分区操作提前,提升数仓ETL的效率。