Hi! If you have last-state upgrade mode configured it may happen that the allowNonRestoredState config is ignored by Flink (as the last-state upgrade mechanism somewhat bypasses the regular submission).
Worst case scenario, you can suspend the deployment, manually record the last checkpoint/savepoint path. Then delete the FlinkDeployment and recreate it with the initialSavepointPath set to your checkpoint. Cheers, Gyula On Thu, Oct 13, 2022 at 7:36 AM Yaroslav Tkachenko <yaros...@goldsky.com> wrote: > Hey everyone, > > I'm trying to redeploy an application using a savepoint. The new version > of the application has a few operators with new uids and a few operators > with the old uids. I'd like to keep the state for the old ones. > > I passed the allowNonRestoredState flag (using Apache Kubernetes Operator > actually) and I can confirm that > "execution.savepoint.ignore-unclaimed-state" is "true" after that. > > However, the application still fails with the following exception: > > "java.lang.IllegalStateException: Failed to rollback to > checkpoint/savepoint s3p://<REDACTED>. Cannot map checkpoint/savepoint > state for operator d9ea0f9654a3395802138c72c1bfd35b to the new program, > because the operator is not available in the new program. If you want to > allow to skip this, you can set the --allowNonRestoredState option on the > CLI." > > Is there a situation where allowNonRestoredState may not work? Thanks. >
