[
https://issues.apache.org/jira/browse/FLINK-40742?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
ASF GitHub Bot updated FLINK-40742:
-----------------------------------
Labels: pull-request-available (was: )
> Pin the JDBC Table sink's changelog behaviour with tests before the Sink V2
> migration
> -------------------------------------------------------------------------------------
>
> Key: FLINK-40742
> URL: https://issues.apache.org/jira/browse/FLINK-40742
> Project: Flink
> Issue Type: Technical Debt
> Components: Connectors / JDBC
> Affects Versions: jdbc-5.0.0
> Reporter: Martijn Visser
> Assignee: Martijn Visser
> Priority: Major
> Labels: pull-request-available
>
> `JdbcDynamicTableSink`'s changelog path is covered by two end-state
> assertions (`testUpsert`, `testReadingFromChangelogSource`); no test
> references `TableBufferReducedStatementExecutor`,
> `TableInsertOrUpdateStatementExecutor`, `TableBufferedStatementExecutor` or
> `TableSimpleStatementExecutor`. FLINK-37987 will replace the legacy
> `SinkFunction` path with a Sink V2 writer; FLINK-26638 (Elasticsearch,
> reverted in 1.15 because a legacy behaviour was not carried over) is the
> reason to pin the behaviour first: reduce by key, upsert-before-delete order,
> retry replay, object-reuse copy, the fallback branch on dialects without a
> native upsert, and `sink.parallelism` above the input parallelism. The
> executor fakes extend `DerbyDialect` and move with FLINK-40730.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
