I wanted to pass this along incase other SDK based SQL Tap developers might have run into this. I was seeing a database session per Stream. The sessions would stay until the

meltano run

was complete. I found that if you add a

tap_connector

property in your Tap's class in `tap.py`:

default_connector_class = mssqlConnector
    _tap_connector: SQLConnector = None

    @property
    def tap_connector(self) -> SQLConnector:
        """The connector object.

        Returns:
            The connector object.
        """
        if self._tap_connector is None:
            self._tap_connector = self.default_connector_class(dict(self.config))
        return self._tap_connector

Then you copy over

catalog_dict

property and

discover_streams

method from the

SQLTap

class in

tap_base.py

. Edit

catalog_dict

to use the new

self.tap_connector

and

discover_streams

to pass

connector=self.tap_connector

to the

default_stream_class

instance as it is appended. This will look like this:

self.default_stream_class(
		tap=self,
		catalog_entry=catalog_entry,
		connector=self.tap_connector
	)

This allows all the Streams to utilize one SQLAlchemy connection pool. When doing test extracts of databases with 80 tables or more I only see a single session utilized. I just merged this change to tap-mssql--buzzcutnorman if you want to see a complete example. If you give this a try in your own SDK based SQL Tap please let me know what the result was, good or bad.