Camus mainly consists of two tasks:
- Reading data from Kafka: Camus acts as a consumer while reading data from Kafka, and it requires the message decoder class, which can be used to read data from Kafka topics. The decoder class must implement com.linkedin.Camus.coders.MessageDecoder class. A few default decoders are available, such as StringMessageDecoder.
- Writing data to HDFS: Camus then writes data read from Kafka into HDFS. To write data into HDFS, it must have a record writer class. The record reader class must implement com.linkedin.Camus.etl.RecordWriterProvider.
Running Camus requires the Hadoop cluster. As discussed, Camus is nothing but a map-reduce job, and it can be run using normal Hadoop jobs as in the following ...