[
https://issues.apache.org/jira/browse/FLINK-37108?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Ruan Hang updated FLINK-37108:
------------------------------
Component/s: Tests
> Source/Sink test suites test invalid recovery
> ---------------------------------------------
>
> Key: FLINK-37108
> URL: https://issues.apache.org/jira/browse/FLINK-37108
> Project: Flink
> Issue Type: Improvement
> Components: Tests
> Affects Versions: 2.0.0, 1.20.0, 1.19.1, 2.0-preview
> Reporter: Arvid Heise
> Assignee: Arvid Heise
> Priority: Major
> Labels: pull-request-available
> Fix For: 2.0.0, 1.19.2, 1.20.1
>
>
> The source and sink test suites have a couple of tests that take a snapshot
> and then recover from it with lower or higher parallelism. However, while
> taking the snapshot they use the `–drain` version
> (advanceToEndOfEventTime=true), which should only be used when terminating
> the pipeline for good. In particular, all operators receive `endInput()` so
> they should finalize their state. The tests, however, add more records to the
> inputs, such that the `endInput` contract is violated.
> The solution is simple: just use advanceToEndOfEventTime=false.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
