I have started experimenting with the couchbase-spark-connector and found it in many ways really simple and nice to connect to Spark.
However, as my application allows to dynamically add and remove connections to data sources, in many ways it has some drawbacks, the worst one being that the couchbase-spark-connector requires me to specify at startup time the available nodes and - even worse - the bucket names.
I tried many things to get around this, but found so far the only way to work to change the connector’s code. Why not allow to add this information as (optional) parameter to the read command? Or am I overlooking something?