Data engineering·Time and correctness in streams
you needed to deploy a new version of the job, so you took a snapshot deliberately, stopped it, and started the new code from that snapshot.
Savepoint
Also calledmanual snapshot, upgrade snapshot
A checkpoint taken on purpose and kept, so a job can be stopped and restarted from it — usually across a code change, a scale change or a cluster migration. It is the streaming equivalent of a deploy artefact, and the constraint it creates is real: your new code has to be able to read the old state, so operator identity and state schema become things you cannot rename casually.