Until now, we used a mysqldump wrapper to save all databases.

We've some issues:

• Ideally, we need to iterate on databases with one file par db
• But in some cases, like for Phabricator, we've 15G db with tables we'd like to ignore, e.g. storageblob, now we've D3730 to use files
  • A backup at table level is needed in that case with ignore rules

P374 offers 3 versions of a script based on those requirements:

1. Procedural, difficult to maintain or extend version
2. Clean architecture version
3. Async version using asyncio (*)

(*) https://docs.python.org/3/library/asyncio.html

Plan is to write our tool, perhaps based on the functions generated by GPT if they can help.