mimaison commented on a change in pull request #11773:
URL: https://github.com/apache/kafka/pull/11773#discussion_r824001349
##########
File path:
connect/api/src/main/java/org/apache/kafka/connect/source/SourceConnector.java
##########
@@ -28,4 +30,46 @@
protected SourceConnectorContext context() {
return (SourceConnectorContext) context;
}
+
+ /**
+ * Signals whether the connector supports exactly-once delivery guarantees
with a proposed configuration.
+ * Developers can assume that worker-level exactly-once support is enabled
when this method is invoked.
+ *
+ * <p>For backwards compatibility, the default implementation will return
{@code null}, but connector developers are
+ * strongly encouraged to override this method to return a non-null value
such as
+ * {@link ExactlyOnceSupport#SUPPORTED SUPPORTED} or {@link
ExactlyOnceSupport#UNSUPPORTED UNSUPPORTED}.
+ *
+ * <p>Similar to {@link #validate(Map) validate}, this method may be
called by the runtime before the
+ * {@link #start(Map) start} method is invoked when the connector will be
run with exactly-once support.
Review comment:
Potentially controversial idea: Have you considered always calling
`start()` first? That way the connector should have computed its configuration
and should be able to easily handle `exactlyOnceSupport()` and
`canDefineTransactionBoundaries()`. Obviously `start()` may cause the connector
to do some work before the runtime stops it in case of an invalid
configuration. But since the proposed javadoc hinted it could be called first
anyway, it's not making first worst
The configuration validation logic is getting more and more complex and
showing the limits of the current APIs. I wonder if having a `configure(Map)`
method would help us.
--
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]
