Hey, Can you guys please help me understand how state recovery works in flink? I'm having a 1.16.2 flink deployment running with 2 JobManagers and 1 taskManager. So if main JobManager dies - second will continue the work. But what if taskManager dies? Will it restore processing from latest checkpoint or from latest savepoint?

Hey! AFAIK, Flink will never use savepoint for restoring automatically. That’s something you specify when you deploy a new job. If a JobManger or TaskManager dies it should restart and use the latest checkpoint to restore.

So, let's say, if I had 100 items processed from kafka after last checkpoint and job failed - it will reprocess these 100 items again?

Yep