Hi,

Want to hear your opinions regarding the right way to handle Kafka
dependency in the Kafka connector. (Related to SQOOP-1852)

Currently, Kafka doesn't have a client module, so to run as a producer
- the connector needs the entire Kafka jar (and its not tiny).

While reviewing the connector, Abe suggested that we may want to add
Kafka dependency as "provided", the way we add Hadoop dependency.

Another option is to treat Kafka jar the way we do JDBC drivers and
ask users to provide it themselves.

Personally, I think that since Kafka is required for this connector to
work, and since its unlikely that Sqoop2 will be installed on a node
with Kafka, we should make life easier on everyone and package this
jar.

Any thoughts about the issue?

Gwen

Reply via email to