Flume+kafka+hdfs detailed explanation

Flume architecture diagram

Single node flume configuration

Flume-1.4.0 starts flume

Bin/flume-ng agent-- conf. / conf-f conf/flume-conf.properties-Dflume.root.logger=DEBUG,console-n agent

-n represents the name of the agent in the configuration file

Agent.sources = r1agent.sinks = s1agent.channels = c1agent.sources.r1.channels = c1agent.sinks.s1.channel = c1#Describe/configure the sourcesagent.sources.r1.type = execagent.sources.r1.command = tail-F / home/flume/loginfo#Use a channel which buffers events in memoryagent.channels.c1.type = memoryagent.channels.c1.capacity = 1000 # Eventagent.channels.c1.transactionCapacity = 100agent.sinks.s1.type = logger

Flume-1.4.0 + kafka-0.7.2+hdfs flume configuration

Agent.sources = r1agent.sinks = s_kafka s_hdfsagent.channels = c_kafka c_hdfsagent.sources.r1.channels = c_kafka c_hdfsagent.sources.r1.type = exec# the following script tail a log agent.sources.r1.command = tail-F / home/flume/loginfoagent.channels.c_kafka.type = memoryagent.channels.c_hdfs.type = memoryagent.sinks.s_kafka.type = com.sink.FirstkafkaSinkagent.sinks.s_kafka.channel = c_kafka#kafka needs to connect to zk Write broker data agent.sinks.s_kafka.zkconnect = localhost:2181agent.sinks.s_kafka.topic = testagent.sinks.s_kafka.serializer.class = kafka.serializer.StringEncoderagent.sinks.s_kafka.metadata.broker.list = localhost:9092 # configuration file server.propertiesagent.sinks.s_kafka.custom.encoding = UTF-8agent.sinks.s_hdfs.type = hdfsagent.sinks.s_hdfs.channel = c_hdfs# default port 8020agent.sinks.s_hdfs.hdfs.path = hdfs://localhost:9000/root/sourceagent.sinks.s_hdfs.hdfs.filePrefix = events-agent.sinks.s_hdfs.hdfs.fileType = DataStreamagent.sinks.s_hdfs.hdfs.writeFormat = Textagent.sinks.s_hdfs.hdfs.rollCount = 30 # reach a certain numerical record generation file agent.sinks.s_hdfs.hdfs.rollSize = 0agent.sinks.s_hdfs.hdfs.rollInterval = 0agent.sinks.s_hdfs.hdfs.useLocalTimeStamp = trueagent.sinks.s_hdfs.hdfs .idleTimeout = 51agent.sinks.s_hdfs.hdfs.threadsPoolSize = 2

Flume built-in channel,source,sink summary http://www.iteblog.com/archives/948