This project is used for flume-ng to communicate with kafka 0.7,2.
agent_log.sinks.kafka_all.type = com.vipshop.flume.sink.kafka.KafkaSink
agent_log.sinks.kafka_all.channel = all_channel
agent_log.sinks.kafka_all.zkconnect = 127.0.0.1:2181
agent_log.sinks.kafka_all.topic = all
agent_log.sinks.kafka_all.batchsize = 200
agent_log.sources.kafka.type = com.vipshop.flume.source.kafka.KafkaSource
agent_log.sources.kafka.channels = internal_channel
agent_log.sources.kafka.zkconnect = 127.0.0.1:2181
agent_log.sources.kafka.batchsize = 200
agent_log.sources.kafka.topic = all
agent_log.sources.kafka.group = hdfs
I have run performance test. One is tail file to kafka and the other is fetch message from kafka. In test I used memory channel in flume, both kafka sink and kafka source can reach 150,000 message per second (when I added pressure to 200,000 message per second, memory channel would have pending message). While the performance hit this awesome performace, I think this plugin for flumg-ng can hold most condition.
In fact I'm a newbie in Java. I have learnt a lot from flumg-ng-rabbitmq. Thanks to Jeremy Custenborder.