Hi, I recently had to do a code update of a long running Flink Stream job (1.3.2) and on the restart from the savepoint I had to deal with:
java.lang.IllegalStateException: Could not initialize keyed state backend. Caused by: java.io.InvalidClassException: lab.vardata.events.Event; local class incompatible: stream classdesc serial VersionUID = 8728793377341765980, local class serialVersionUID = -4253404384162522764 because I have changed a method used to convert the Event to a Cassandra writable Tuple (in particular, I changed the return type from Tuple10 to Tuple11, after adding a field). I reverted those changes back since it wasn't much of a problem per se. Now, I understand the root cause of this issue and I wanted to ask if there are any "best practices" to prevent this kind of issues, without losing the state of the job, because of restarting it from the very beginning. -- Federico D'Ambrosio
