Hi Team, We are trying to use savepoints with Flink Operator Requirement : • If Task Manager Restarts due to error, it should pick the last Savepoint state • If Job Manager Restarts due to error, it should pick the last Savepoint state • If we manually delete and start the job again, it should pick the last Savepoint In all the above cases our state.savepoint.dir stays the same We are savepointing to S3 directory which is persistent, so expect the state to be always present there unless cleaned up by the job as per history configuration. Can someone help with the configurations to achieve this with Flink Operator ? Also confirm if this is possible to achieve with Flink Operator or we need to implement any pre, post jobs to save and restore from savepoint ? Appreciate any inputs on this. Thanks

Hi Dheeraj, Please refer to this doc https://nightlies.apache.org/flink/flink-kubernetes-operator-docs-release-1.5/docs/custom-resource/reference/#upgrademode

Hi @Vipul Jindal, that for sharing. We are already using the upgrade mode as Savepoint, yet we see when the job restarts, it does not pick up from the previous savepoint

Ahh, I dont have much Ida about automatic restores from a savepoint as we use last-state which takes the last n checkpoints.