This sounds a bit unusual. Why would someone propose supporting Starrocks’ catalog within Starrocks itself?
However, since your company currently has multiple Starrocks clusters with different versions, such as version 2.5 and version 3.1, you encounter issues when accessing tables from the 2.5 version in the 3.1 version due to some historical reasons. Currently, your approach is to create a duplicate table in the 3.1 version for the tables in the 2.5 version and use data synchronization to write data. You have also attempted using the JDBC catalog approach but found that it resulted in the loss of information such as table primary key patterns or indexes, which leads to suboptimal queries.
Therefore, you would like to initiate a discussion in the Starrocks forum to explore the possibility of having a Starrocks catalog on Starrocks. You welcome everyone to participate in the discussion.