案例八：多flume案例

画图理解

配置文件编写

第一个服务器

[root@node1 one]# cat fileToFlume.conf 
one.sources = r1
one.sinks = k1
one.channels = c1
 
one.sources.r1.type = exec
one.sources.r1.command = tail -F /opt/data/one/one.txt
 
one.sinks.k1.type = avro
one.sinks.k1.hostname = node1
one.sinks.k1.port = 40000
 
one.channels.c1.type = memory
one.channels.c1.capacity = 1000
one.channels.c1.transactionCapacity = 100
 
one.sources.r1.channels = c1
one.sinks.k1.channel = c1

第二个服务器

[root@node1 two]# cat fileToFlume.conf 
two.sources = r1
two.sinks = k1
two.channels = c1
 
two.sources.r1.type = exec
two.sources.r1.command = tail -F /opt/data/two/two.txt
 
two.sinks.k1.type = avro
two.sinks.k1.hostname = node1
two.sinks.k1.port = 40000
 
two.channels.c1.type = memory
two.channels.c1.capacity = 1000
two.channels.c1.transactionCapacity = 100
 
two.sources.r1.channels = c1
two.sinks.k1.channel = c1

聚合服务器

[root@node1 aggre]# cat fileToHdfs.sh 
# 给flume采集进程agent起了一个别名  a1  
# 定义flume进程中有几个source 以及每一个source的别名
aggre.sources = r1
aggre.sinks = k1
aggre.channels = c1
 
# 定义flume进程中source连接到数据源信息
aggre.sources.r1.type = avro
aggre.sources.r1.bind = node1
aggre.sources.r1.port = 40000
 
# sink info
aggre.sinks.k1.type = hdfs
#指定采集的数据存储到HDFS的哪个目录下
aggre.sinks.k1.hdfs.path = hdfs://node1:9000/log/
#上传文件的前缀
aggre.sinks.k1.hdfs.filePrefix = web-
#上传文件的后缀
aggre.sinks.k1.hdfs.fileSuffix = .log
##是否按照时间滚动文件夹
aggre.sinks.k1.hdfs.round = false
##是否使用本地时间戳
aggre.sinks.k1.hdfs.useLocalTimeStamp = true
##积攒多少个Event才flush到HDFS一次
aggre.sinks.k1.hdfs.batchSize = 1000
##设置文件类型，可支持压缩
aggre.sinks.k1.hdfs.fileType = DataStream
##多久生成一个新的文件
aggre.sinks.k1.hdfs.rollInterval = 3600
##设置每个文件的滚动大小
aggre.sinks.k1.hdfs.rollSize = 134217700
##文件的滚动与Event数量无关
aggre.sinks.k1.hdfs.rollCount = 0
##最小冗余数
aggre.sinks.k1.hdfs.minBlockReplicas = 1
 
aggre.channels.c1.type = memory
aggre.channels.c1.capacity = 1000
aggre.channels.c1.transactionCapacity = 100
 
aggre.sources.r1.channels = c1
aggre.sinks.k1.channel = c1