Migrating stateful Flink jobs
When you run Flink application with state, you must stop the Flink jobs with a savepoint, so you can restart them from the exact point on your upgraded Data Hub clusters. Based on the version of your original cluster, you can provide the savepoint path either in HDFS or S3.
- If the version of the original cluster was 7.2.9 or lower, you need to provide the savepoint path in HDFS.
- If the version of the original cluster was 7.2.10 or higher, you need to provide the savepoint path in S3.