We should re-allow Foreign Keys again between clusterwide and cell database tables?
As part of the new [Cells 1.0] architecture, gitlab_main_clusterwide
tables don't mean the same thing as clusterwide
in the old Cells architecture. Which means that they are not shared by cells, and are not accessed directly by more than 1 cell.
Each Cell has its own gitlab_main_clusterwide
, gitlab_main_cell
tables.
gitlab_main_clusterwide
tables will remain on the same database cluster where gitlab_main_cell
, inside each cell. And because of this change, we have allowed again the cross joins and cross database modification between these two tables.
Should we also allow FKs between these two table schemas? or do we see a problem moving forward in Cells 1.5/Cells 2.0 when we users can interact with more than 1 organization?
If can allow the FKs again, we can do a similar MR to this one: !145669 (diffs) but for FKs instead. We can change this issue to work issue, instead of discussion
Update
It was decided in an Internal Slack Thread to re-allow FKs between the two gitlab schema tables, because they will stay on the same database for all the Cells.