Hey folks! 👋
I’m using Flink 1.19 with the Kafka connector (flink-connector-kafka:3.3.0) and have noticed some unexpected behavior.
During checkpoints, the committed offsets are correctly synchronized with the Kafka consumer group. However, when invoking stop with savepoint, the latest committed offsets are stored in the savepoint state but are not synced to the Kafka consumer group.
As a result:
• The consumer group’s last committed offsets reflect the previous completed checkpoint.
• Meanwhile, the savepoint contains the actual latest offsets.
Has anyone encountered this issue before? Any insights or workarounds would be greatly appreciated!