Deep Dive deadline storage method

[ferruzzi]

Determine storage method (in the DAG object or in a DB table or ??? )

• Must be able to sort/search by "need-by" and remove by "dagrun_id"
• Must be able to allow alternative remove options for task_id or other options for future work

[ferruzzi]

Current thought:

A new table in the DB which has a 1:1 mapping for identifier:deadline, where identifier can be a dagrun_id, task_id, etc. and a query which returns the deadlines filtered to only return the ones which have passed

[ferruzzi]

If we are only storing dagrun_id and task_id in the identifier column then it's easy to infer what type of id it is from the format, but does every possible future identifier type have a name format where we can infer what it is from the pattern? Presumably not? So we likely need a type column as well just to be safe?