hackergin commented on PR #7914:
URL: https://github.com/apache/paimon/pull/7914#issuecomment-4504581933
State compatibility / upgrade risk
This fix changes the resulting job graph topology for any existing user
who has scan.parallelism set or relies on inferred parallelism:
- Before: source and downstream operators are chained via forward edges,
all running at source parallelism.
- After: the planner inserts a rebalance edge between source and
downstream, and operator chaining is broken.
For users upgrading from an older Paimon version with an existing Flink
savepoint/checkpoint, this is a breaking topology change:
- The operator IDs / chain hashes of downstream operators will change, so
state restore from savepoint will fail (or silently drop state) unless users
explicitly set uids on
every operator — which most SQL users don't.
- Even watermark/state alignment behavior of downstream stateful operators
may differ once the chain is broken.
This needs to be called out explicitly in the PR description / release
notes as a breaking change for stateful streaming jobs.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
