利用Apache Flume 读取JMS 消息队列消息。并将消息写入HDFS，flume agent配置例如以下：
flume-agent.conf
#name the components on this agent
agentHdfs.sources = jms_source
agentHdfs.sinks = hdfs_sink
agentHdfs.channels = mem_channel
# Describe/configure the source
agentHdfs.sources.jms_source.type = jms
# Bind to all interfaces
agentHdfs.sources.jms_source.initialContextFactory = org.apache.activemq.jndi.ActiveMQInitialContextFactory
agentHdfs.sources.jms_source.connectionFactory = ConnectionFactory
agentHdfs.sources.jms_source.destinationName = BUSINESS_DATA #AMQ queue
agentHdfs.sources.jms_source.providerURL = tcp://hadoop-master:61616
agentHdfs.sources.jms_source.destinationType = QUEUE
# Describe the sink
agentHdfs.sinks.hdfs_sink.type = hdfs
agentHdfs.sinks.hdfs_sink.hdfs.path hdfs://hadoop-master/data/flume/%Y-%m-%d/%H
agentHdfs.sinks.hdfs_sink.hdfs.filePrefix = %{hostname}/events-
agentHdfs.sinks.hdfs_sink.hdfs.maxOpenFiles = 5000
agentHdfs.sinks.hdfs_sink.hdfs.batchSize= 500
agentHdfs.sinks.hdfs_sink.hdfs.fileType = DataStream
agentHdfs.sinks.hdfs_sink.hdfs.writeFormat =Text
agentHdfs.sinks.hdfs_sink.hdfs.rollSize = 0
agentHdfs.sinks.hdfs_sink.hdfs.rollCount = 1000000
agentHdfs.sinks.hdfs_sink.hdfs.rollInterval = 600
agentHdfs.sinks.hdfs_sink.hdfs.useLocalTimeStamp = true
# Use a channel which buffers events in memory
agentHdfs.channels.mem_channel.type = memory
agentHdfs.channels.mem_channel.capacity = 1000
agentHdfs.channels.mem_channel.transactionCapacity = 100
# Bind the source and sink to the channel
agentHdfs.sources.jms_source.channels = mem_channel
agentHdfs.sinks.hdfs_sink.channel = mem_channel