Flume作业nc-memory-logger|HDFS

Flume日志采集配置

最新推荐文章于 2024-08-07 11:13:35 发布

转载最新推荐文章于 2024-08-07 11:13:35 发布 · 118 阅读

本文介绍如何使用Flume从指定网络端口采集日志，并将其输出到控制台及HDFS中。具体包括配置源、接收器、通道及处理器等组件，实现日志数据的收集与存储。

需求1：
从指定的网络端口上采集日志到控制台输出和HDFS

负载算法

故障转移： 可以指定优先级，数字越大越优先

a1.sinkgroups.g1.processor.type = failover

a1.sinkgroups = g1
a1.sinkgroups.g1.sinks = k1 k2
a1.sinkgroups.g1.processor.type = failover
a1.sinkgroups.g1.processor.priority.k1 = 5
a1.sinkgroups.g1.processor.priority.k2 = 10
a1.sinkgroups.g1.processor.maxpenalty = 10000

全部轮询

a1.sinkgroups.g1.processor.type = load_balance

#从指定的网络端口上采集日志到控制台输出和HDFS

a1.sources = r1
a1.sinks = k1 k2 
a1.channels = c1
# Describe/configure the source
a1.sources.r1.type = netcat
a1.sources.r1.bind = 0.0.0.0
a1.sources.r1.port = 44444
# Describe the sink
a1.sinkgroups = g1
a1.sinkgroups.g1.sinks = k1 k2
a1.sinkgroups.g1.processor.type = load_balance
a1.sinks.k1.type = logger
a1.sinks.k2.type = hdfs
a1.sinks.k2.hdfs.path = hdfs://192.168.0.129:9000/user/hadoop/flume
a1.sinks.k2.hdfs.batchSize = 10
a1.sinks.k2.hdfs.fileType = DataStream
a1.sinks.k2.hdfs.writeFormat = Text
# Use a channel which buffers events in memory
a1.channels.c1.type = memory
a1.channels.c1.capacity = 1000
a1.channels.c1.transactionCapacity = 100
# Bind the source and sink to the channel
a1.sources.r1.channels = c1
a1.sinks.k1.channel = c1
a1.sinks.k2.channel = c1

检查logger输出：

2018-08-10 18:58:39,659 (lifecycleSupervisor-1-3) [INFO - org.apache.flume.source.NetcatSource.start(NetcatSource.java:169)] Created serverSocket:sun.nio.ch.ServerSocketChannelImpl[/0:0:0:0:0:0:0:0:44444]
2018-08-10 18:59:17,723 (SinkRunner-PollingRunner-LoadBalancingSinkProcessor) [INFO - org.apache.flume.sink.LoggerSink.process(LoggerSink.java:94)] Event: { headers:{} body: 7A 6F 75 72 63 20 6F 6B 0D                      zourc ok. }
2018-08-10 19:00:35,744 (SinkRunner-PollingRunner-LoadBalancingSinkProcessor) [INFO - org.apache.flume.sink.LoggerSink.process(LoggerSink.java:94)] Event: { headers:{} body: 61 73 64 66 0D                                  asdf. }
2018-08-10 19:00:35,774 (SinkRunner-PollingRunner-LoadBalancingSinkProcessor) [INFO - org.apache.flume.sink.hdfs.HDFSDataStream.configure(HDFSDataStream.java:58)] Serializer = TEXT, UseRawLocalFileSystem = false
2018-08-10 19:00:36,086 (SinkRunner-PollingRunner-LoadBalancingSinkProcessor) [INFO - org.apache.flume.sink.hdfs.BucketWriter.open(BucketWriter.java:234)] Creating hdfs://192.168.0.129:9000/user/hadoop/flume/FlumeData.1533942035775.tmp

检查hdfs输出：

[hadoop@hadoop001 flume]$ hdfs dfs -text  hdfs://192.168.0.129:9000/user/hadoop/flume/*
18/08/10 19:14:23 WARN util.NativeCodeLoader: Unable to load native-hadoop library for your platform... using builtin-java classes where applicable
zourc1
2
3
4
5
6
7
8
9
10

来自 “ ITPUB博客 ” ，链接：http://blog.itpub.net/31441024/viewspace-2199830/，如需转载，请注明出处，否则将追究法律责任。

转载于:http://blog.itpub.net/31441024/viewspace-2199830/