Hi, We are running a flink job in session mode and if one TaskManager pod within a Flink job fails, the overall Flink job is restarting. Is there a way that flink job can continue to run without interruption incase of any taskmanager pod failure ?

No. If a TaskManager running parts of a Flink job fails, the job needs to be restarted (if you have a job consisting of multiple disconnected DAGs, only the DAG affected by the failure will restart). A full restart is required to reset the state to the last successfully completed checkpoint

@rmetzger Thank you for the response 🙂 One qq on RocksDB : We are using RocksDB as the state backend for our Flink application. Is it possible to extract data from RocksDB and save it as a CSV file? Requirement is to do data reconciliation between what we transmit to our Elastic Cluster and what Flink processes.

You can use the state processor api to inspect data in checkpoints and savepoints: https://nightlies.apache.org/flink/flink-docs-master/docs/libs/state_processor_api/