多个子节点收集日志-主节点上传到HDFS

标签：inf   bsp   client   write   pac   poi   ons   idle   bin   

Master:

----------------------

#MasterAgent
MasterAgent.channels = c1
MasterAgent.sources = s1
MasterAgent.sinks = k1

#MasterAgent Avro Source
MasterAgent.sources.s1.type = avro
MasterAgent.sources.s1.bind = 192.168.192.128
MasterAgent.sources.s1.port = 4444

##MasterAgent FileChannel
#MasterAgent.channels.c1.type = file
#MasterAgent.channels.c1.checkpointDir = /home/hadoop/data/spool/checkpoint
#MasterAgent.channels.c1.dataDirs = /home/spool/data
#MasterAgent.channels.c1.capacity = 200000000
#MasterAgent.channels.c1.keep-alive = 30
#MasterAgent.channels.c1.write-timeout = 30
#MasterAgent.channels.c1.checkpoint-timeout=600

#MasterAgent Memory Channel
MasterAgent.channels.c1.type = memory
MasterAgent.channels.c1.capacity = 1000
MasterAgent.channels.c1.transactionCapacity = 100
#MasterAgent.channels.c1.byteCapacityBufferPercentage = 20
#MasterAgent.channels.c1.byteCapacity = 800

# Describe the sink
MasterAgent.sinks.k1.type = hdfs
MasterAgent.sinks.k1.hdfs.path =hdfs://qm1711/flume_data
MasterAgent.sinks.k1.hdfs.filePrefix = %Y_%m_%d
MasterAgent.sinks.k1.hdfs.fileSuffix = .log
MasterAgent.sinks.k1.hdfs.round = true
MasterAgent.sinks.k1.hdfs.roundValue = 1
MasterAgent.sinks.k1.hdfs.roundUnit = minute
MasterAgent.sinks.k1.hdfs.useLocalTimeStamp = true
MasterAgent.sinks.k1.hdfs.fileType=DataStream
MasterAgent.sinks.k1.hdfs.writeFormat=Text
MasterAgent.sinks.k1.hdfs.rollInterval=0
MasterAgent.sinks.k1.hdfs.rollSize=10240
MasterAgent.sinks.k1.hdfs.rollCount=0
MasterAgent.sinks.k1.hdfs.idleTimeout=60
MasterAgent.sinks.k1.hdfs.callTimeout=60000

MasterAgent.sources.s1.channels = c1
MasterAgent.sinks.k1.channel = c1

-------------

Slave:

#SlavesAgent
SlavesAgent.channels = c1
SlavesAgent.sources = s1
SlavesAgent.sinks = k1

#SlavesAgent Spooling Directory Source
SlavesAgent.sources.s1.type = spooldir
SlavesAgent.sources.s1.spoolDir =/home/hadoop/logs/
SlavesAgent.sources.s1.fileHeader = true
SlavesAgent.sources.s1.deletePolicy =immediate
SlavesAgent.sources.s1.batchSize =1000
SlavesAgent.sources.s1.deserializer.maxLineLength =1048576

#SlavesAgent Memory Channel
SlavesAgent.channels.c1.type = memory
SlavesAgent.channels.c1.capacity = 1000
SlavesAgent.channels.c1.transactionCapacity = 100
#SlavesAgent.channels.c1.byteCapacityBufferPercentage = 20
#SlavesAgent.channels.c1.byteCapacity = 800

# SlavesAgent Sinks
SlavesAgent.sinks.k1.type = avro
# connect to CollectorMainAgent
SlavesAgent.sinks.k1.hostname = 192.168.192.128
SlavesAgent.sinks.k1.port = 4444

SlavesAgent.sources.s1.channels =c1
SlavesAgent.sinks.k1.channel = c1

命令：

开启agent:
flume-ng agent -c . -f /home/hadoop/flume-1.7.0/conf/myconf/MasterNode.conf -n MasterAgent -Dflume.root.logger=INFO,console
flume-ng agent -c . -f /home/hadoop/flume-1.7.0/conf/myconf/SlavesNode.conf -n SlavesAgent -Dflume.root.logger=INFO,console
执行要监控的脚本文件
flume-ng avro-client -c . -H localhost -p -F /data.txt

多个子节点收集日志-主节点上传到HDFS

标签：inf   bsp   client   write   pac   poi   ons   idle   bin   

原文地址：https://www.cnblogs.com/pingzizhuanshu/p/9102538.html