With some initial work done on
DM-29593 it is time now to switch to specific task of migrating datasets schema to UUID keys. Few items that need some work:
- we have relatively large number of records in existing repos, tens(s) of millions
- this needs special care for performance, there are couple of possible options for how to perform this migratrion:
- dump everything into CSV, run bulk transforms on those CSVs and import it back into database
- make special mapping table from int dataset_id into UUID and generate new tables by doing "INSERT ... SELECT ... JOIN"
- latter should be cleaner but there is a question of JOIN performance which I want to study first