https://streamacademy.io/tags/savepoints/ Recent content in Savepoints on Data Streaming Academy Hugo en-US Copyright © 2025 - 2026 Irontools Technologies Inc. Thu, 16 Apr 2026 21:18:05 -0700 https://streamacademy.io/tutorial/flink-kafka-consumer-offsets-with-the-state-processor-api/ Wed, 15 Apr 2026 00:00:00 -0700 https://streamacademy.io/tutorial/flink-kafka-consumer-offsets-with-the-state-processor-api/ <h1 id="apache-flink-reading-and-modifying-kafka-consumer-offsets-using-the-state-processor-api">Apache Flink: Reading and Modifying Kafka Consumer Offsets Using the State Processor API</h1> <p><a href="https://nightlies.apache.org/flink/flink-docs-release-2.2/docs/connectors/datastream/kafka/#starting-offset" target="_blank" rel="noopener">Kafka Consumer offset management</a> in the Flink Kafka source is notoriously difficult to understand. The offsets are stored in the Flink state, but they <a href="https://nightlies.apache.org/flink/flink-docs-release-2.2/docs/connectors/datastream/kafka/#consumer-offset-committing" target="_blank" rel="noopener">can also be committed to Kafka</a>. Flink generally uses the offsets stored in state, unless a specific configuration is provided to favor Kafka consumer group offsets.</p> <p>So, understanding what offsets Flink actually stores can be very important for many situations. Maybe you need to rewind them due to a data corruption bug. Or maybe a Flink job started from a savepoint consumes data it&rsquo;s not supposed to, and you&rsquo;d like to understand what a savepoint actually contains. Or maybe you&rsquo;re migrating a job between clusters and need to bootstrap state from scratch.</p>