SEPURI-SAI-KRISHNA commented on PR #13250:
URL: https://github.com/apache/seatunnel/pull/13250#issuecomment-6907380363

   @DanielLeens Thanks, all three are done in `ca6cde4e98`.
   
   **Issue 1.** `Jdbc.md` (EN and ZH) has a note under `primary_keys`: with 
`multi_table_sink_replica` above 1 and `is_exactly_once = false`, each row goes 
to a writer by the first column of the table's primary key, or, when the table 
has none, of `primary_keys` (configured or inherited from a unique key), so the 
changes of one key reach one writer in order. It also states the limits: rows 
of a table without such a key, and all rows with `is_exactly_once = true`, are 
not routed by key; an update that changes the key value can move the record to 
another writer; rows in different writers have no order.
   
   **Issue 2.** The IT empties `sink_unique` before the job and again in 
`finally`, so a failed run on one engine leaves nothing for the next.
   
   **Issue 3.** `createCatalogTableWithUniqueKeyOnName` says the table 
deliberately has no primary key, since the factory then takes `primary_keys` 
from the unique key and a primary key would test the other path. 
`createJdbcSink` says it validates the options and goes through 
`factory.createSink`, as a job submission does.
   
   CI: the rerun of 
[43367554092](https://github.com/SEPURI-SAI-KRISHNA/seatunnel/actions/runs/43367554092)
 on the previous head passed, including the JDK 17 `edge-agent-it` lane that 
had failed (`JobMasterTest` "Node failed to start!", not in the JDBC code). 
`JdbcSinkFactoryTest` 22 and `JdbcSinkTest` 7 pass locally, and `MarkdownTest` 
passes.
   


-- 
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]

Reply via email to