litiliu opened a new pull request, #4608:
URL: https://github.com/apache/flink-cdc/pull/4608
<!--
Licensed to the Apache Software Foundation (ASF) under one
or more contributor license agreements. See the NOTICE file
distributed with this work for additional information
regarding copyright ownership. The ASF licenses this file
to you under the Apache License, Version 2.0 (the
"License"); you may not use this file except in compliance
with the License. You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing,
software distributed under the License is distributed on an
"AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
KIND, either express or implied. See the License for the
specific language governing permissions and limitations
under the License.
-->
## What is the purpose of this pull request?
Fix [FLINK-41313](https://issues.apache.org/jira/browse/FLINK-41313).
When a checkpoint references a Fluss partition that is dropped or expires
before recovery, the pipeline source can retain obsolete splits and block
source progress on repeated metadata requests for missing partition IDs. The
same missing retirement also affects running jobs while their logical table
remains subscribed. This change retires partitions absent from a successful
complete discovery, filters stale restored splits before SDK subscription, and
continues reading surviving partitions from their checkpoint positions.
## Brief change log
- Reconcile assigned/pending partitions and snapshot-consumption bookkeeping
before publishing a complete physical-table inventory to readers, including
newly registered readers.
- Filter restored and late splits against the current inventory; reject
obsolete asynchronous initialization results and returned splits using current
bucket identities.
- Unsubscribe the affected SDK partition/buckets and cancel their
active/queued snapshot scans on the fetcher thread. Use normal finished-split
cleanup and preserve resources and schema state for surviving partitions.
- Keep existing enumerator/split serialization versions and split IDs.
Same-name partition recreation with a new physical ID remains outside this
change.
---
## Verifying this change
Added ITs for partition-scoped SDK cleanup and whole-job checkpoint recovery
after partition deletion, covering `earliest` and `full` startup modes.
## Documentation
- Does this pull request introduce a new feature? No; this is a bug fix.
- If yes, how is the feature documented? Not applicable.
---
##### Was generative AI tooling used to co-author this PR?
- [X] Yes — OpenAI Codex (GPT-6); GitHub Copilot (Claude Opus 5.5, read-only
design discussion).
Generated-by: OpenAI Codex (GPT-6); GitHub Copilot (Claude Opus 5.5)
--
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]
