SEPURI-SAI-KRISHNA commented on issue #10753:
URL: https://github.com/apache/seatunnel/issues/10753#issuecomment-6547035235

   A note on the **JMS / AMQP (generic)** row, which I would like to pick up. 
It is listed as "Partial" with no owner, the tracker notes that the ActiveMQ 
Classic source (#12341) does not complete the generic request, and an earlier 
refresh here described it as "pending scope clarification". I looked into what 
a generic version would need and wrote it up as a design note in #13078. Some 
highlights:
   
   - **Scope:** one `connector-jms` on the standard JMS 1.1 API, with each 
vendor's client jars supplied by the user, as with JDBC drivers. That would 
cover Artemis, IBM MQ, TIBCO EMS and Solace, and AMQP 1.0 brokers through 
Apache Qpid JMS. `connector-activemq` and `connector-rabbitmq` stay as they are.
   - **Why #12341's approach does not carry over:** #12341 acknowledges 
messages one at a time with an ActiveMQ-only mode. Standard JMS acknowledgement 
covers the whole session, so a generic source needs another way to tie delivery 
to checkpoints. The note proposes one transacted session per checkpoint epoch, 
committed only after that checkpoint completes, including the final checkpoint 
when a job stops.
   - **Checked before writing it:** I ran the key behaviour against three 
brokers (ActiveMQ Classic, Artemis, and Artemis over AMQP 1.0 through Qpid JMS) 
on JDK 11 and 17. A negative control shows that the simpler single-session 
approach does lose a message after a failure.
   - **Licence:** the only new dependency would be the Apache Geronimo JMS 1.1 
spec jar, already in the build through `activemq-client`. No vendor client 
would be bundled.
   
   The note ends with four open questions: the scope, the naming, where vendor 
jars should go, and the order of the slices. Feedback on them is very welcome 
on #13078, so the design discussion stays out of this tracker.
   
   If the direction looks right, I would be glad to take this row and implement 
it in two slices, the source first and then the sink. Could the row be recorded 
as claimed by me, in design review under #13078, so the work is not duplicated 
while those questions are open? I will not start coding until they are answered.
   


-- 
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