The cachedb is what actually submits jobs and starts tracking stuff. It should also track reference counts (which dags from which repos called for this execution). That way users can cancel running dags and dml can propagate that cancellation properly. Also, users could delete dags/repos and dml would propagate that too.
We'd need to store it as a doubly linked deal so that we can easily look up "who are the children of this dag" as well as "are there any other valid parents of this dag".
The cachedb is what actually submits jobs and starts tracking stuff. It should also track reference counts (which dags from which repos called for this execution). That way users can cancel running dags and
dmlcan propagate that cancellation properly. Also, users could delete dags/repos and dml would propagate that too.We'd need to store it as a doubly linked deal so that we can easily look up "who are the children of this dag" as well as "are there any other valid parents of this dag".