package streamloader
The entry point of the stream loader library is the StreamLoader class, which requires a KafkaSource and a Sink. Once started it will subscribe to the provided topics and will start polling and sinking records. The sink has to be able to persist records and to look up committed offsets (technically this is optional, but without it there would be no way to provide any delivery guarantees). A large class of sinks are batch based, implemented as RecordBatchingSink. This sink accumulate batches of records using some RecordBatcher and once ready, stores them to some underlying RecordBatchStorage. A common type of batch is file based, i.e. a batcher might write records to a temporary file and once the file is full the sink commits the file to some underlying storage, such as a database or a distributed file system like HDFS.
A sketch of the class hierarchy illustrating the main classes and interfaces can be seen below.
For concrete storage implementations see the clickhouse, hadoop, s3 and vertica packages. They also contain more file builder implementations than just the CsvFileBuilder included in the core library.
- Alphabetic
- By Inheritance
- streamloader
- AnyRef
- Any
- Hide All
- Show All
- Public
- Protected
Package Members
Type Members
- class StreamLoader extends Logging with Metrics
The main stream loading class, given a KafkaSource and some Sink continuously polls the source and sinks records to the sink.
The main stream loading class, given a KafkaSource and some Sink continuously polls the source and sinks records to the sink.
Runs in the active thread and blocks it. When running multiple instances in multiple threads you must provide separate sources and sinks.