Yukang-Lian opened a new pull request, #67265:
URL: https://github.com/apache/doris/pull/67265
### What problem does this PR solve?
Issue Number: None
Related PR: #61143
Problem Summary: A transaction can become visible while part of its publish
work is still incomplete. Schema change and rollup jobs could then send alter
tasks with the partition visible version before enough corresponding base
replicas were ready, causing backends to reject the request because the base
tablet version was behind. Keep local jobs in WAITING_TXN until every tablet
has a quorum of alive, schedule-available, version-complete source replicas.
Treat a missing source-replica mapping as unavailable, build each retry batch
atomically, and preserve the existing cloud path.
### Release note
Schema change and rollup jobs now wait for a ready base-replica quorum
before dispatching alter tasks.
### Check List (For Author)
- Test: Unit Test and Regression test
- Focused SchemaChangeJobV2Test and RollupJobV2Test coverage
- Docker regression for one ready replica followed by quorum recovery
- Behavior changed: Yes. Local alter jobs can remain in WAITING_TXN until a
ready base-replica quorum exists.
- Does this need documentation: No
Co-authored-by: Siyang Tang <[email protected]>
--
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]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
