Apache Flink 1.7–1.14 Release Highlights and Feature Evolution

Flink 1.7

Introduced Scala 2.12 support, Exactly‑once S3 StreamingFileSink, MATCH_RECOGNIZE in Streaming SQL, Temporal Tables and Joins, a versioned REST API, Kafka 2.0 connector, local recovery, and removal of the legacy mode.

Flink 1.8

Added Schema Evolution, TTL‑based state cleanup, user‑defined functions in SQL, RFC‑compliant CSV format, new KafkaDeserializationSchema, and several connector and API improvements.

Flink 1.9

Featured fine‑grained batch job recovery, State Processor API, stop‑with‑savepoint, a redesigned Web UI, Blink‑based SQL planner preview, and extensive Table API/SQL enhancements including DDL support and temporal table joins.

Flink 1.10

Major upgrade with over 200 contributors, improved memory management, unified job submission logic, native Kubernetes (beta) integration, production‑ready Hive integration, expanded PyFlink UDF support, and numerous connector and optimizer updates.

Flink 1.11

Made Blink planner the default, introduced unaligned checkpoints, a new Source API, CDC support in Table & SQL, JDBC catalog, enhanced Hive real‑time warehousing, and significant stability and production‑ready features such as application mode and improved checkpointing.

Flink 1.12

Unified DataStream API with batch execution mode, new Data Sink API, Kubernetes‑based HA, metadata handling in SQL connectors, upsert‑Kafka connector, temporal table joins, and extensive Python API enhancements including stateful DataStream operations and UDAFs.

Flink 1.13

Introduced window table‑valued functions, declarative source/sink registration, a new code generator, removal of the old planner, Python DataStream API performance improvements, LoopBack mode for local execution, and many Table API/SQL refinements.

Flink 1.14

Focused on unified streaming‑batch processing, reactive scaling, performance diagnostics (back‑pressure, CPU flame graphs, state access latency), checkpointing for finite streams, state‑backend switching with savepoints, fine‑grained resource management, and further SQL/Table API enhancements such as session windows and improved code generation.

Key Code Snippets

./bin/flink run -d -e kubernetes-session -Dkubernetes.cluster-id= examples/streaming/WindowJoin.jar

CREATE TABLE kafka_table (
  id BIGINT,
  name STRING,
  event_time TIMESTAMP(3) METADATA FROM 'timestamp',
  headers MAP METADATA
) WITH (
  'connector' = 'kafka',
  'topic' = 'test-topic',
  'format' = 'avro'
);

StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
env.setRuntimeMode(RuntimeMode.BATCH);