zhisheng17/flink-learning

This repository, written mostly in Chinese, is a long-running learning project built around Apache Flink, a system for processing streams of data in real time. The author, zhisheng17, gathered code samples, blog posts, and source code walkthroughs in one place so that readers can study Flink from a beginner level up through performance tuning and internal mechanics. The project is paired with a paid column called Flink in Action and Performance Tuning, and the code in the repo is what those articles refer to. The code itself is Java. According to the change log inside the README, the project has been upgraded several times: to Flink 1.10 in early 2020, to 1.13.2 in August 2021, and to 1.14.2 in December 2021. Older versions are kept on separate branches so readers studying earlier API shapes can still find working code. The build uses Maven, and the README suggests adding an Aliyun mirror to settings.xml for users in China before running the package command. The samples cover a wide range of integrations rather than one single demo. The topics list and the article index include Kafka, MySQL, HBase, Elasticsearch, Redis, Cassandra, InfluxDB, ClickHouse, RocketMQ, Flume, HDFS, and Loki, and many sample programs follow the same pattern: read data from Kafka and write it into one of those other systems. Other articles cover Flink concepts such as windows, watermarks, event time, parallelism, slots, JobManager high availability, and checkpoints. A second series in the README is a source code analysis, walking through how Flink starts up in local and standalone modes, how JobManager and TaskManager interact, how SubmitJob is handled, and how checkpointing, serialization, memory management, and metrics are implemented internally. There is also a paper folder with references about stream processing engines, and a books section the author added in 2022. The README is a long index page rather than a usage guide, and many of the linked articles sit behind the author's blog, WeChat account, or a paid knowledge platform. The full README is longer than what was shown.