Comment 2 for bug 1741161

This question needs to be divided into two stages, and I now have only the experience of mongodb and mariadb galera, which are two different forms of clustering.
mariadb galera can perform database and user operations on any node in the cluster, but mongodb can only be performed on mongos nodes.
1. Into _find_instance not a wrong amendment, but the feeling is not perfect, can be used to transition.
2. We can enhance this functionality by parsing the datastore of the cluster when it is determined that the cluster is passing in and then looking for a suitable database schema for each of these operations but this requires having all kinds of clusters supported by trove experience of