Ability to replay requests (Access and Erasure)

[adriaaaa]

Description

In the scenario of where there is a failure, we will need to support replaying the request, both at the entire Subject Request level and the individual datastore level.

“replaying” = starting the subject request from any point in it’s workflow regardless of if it failed or not.

[pattisdr]

I misunderstood this ticket and assumed we meant "restart from failure" here, but @seanpreston tells me this ticket is primarily for allowing the user to start from an arbitrary point in the graph where there was no failure. Further product discussion is needed here to refine expected behavior.

Restart from failure has been reticketed here: #574

[adriaaaa]

Let's clarify this tomorrow @pattisdr . Your comment on #863 makes me think that we don't have the abililty to restart from failure, but on this ticket it says that "restart from failure" was ticketed on #574 .

[pattisdr]

We can "restart from failure" - by that, I mean if a graph fails, when we restart, only the failed collection and any collections that didn't run the first time will run. For example, if we have collections A, B, and C, and A runs, but then B fails, "restart from failure" will run graphs B and C only.

We can "resume from pause" - in that a graph can exit when a collection indicates it needs manual data and we can resume from that collection. For example, say we have collections A, B, and C, and A runs, but B pauses the graph because it is waiting on manual data and C hasn't gotten a chance to run yet. The user can hit an API endpoint with the manual data to resume, which will get passed into collection B and then collection C will run. A will not run again.

We do not support "replay" (using Sean's definition) where a user can arbitrarily say "I want to rerun collection A again" even though A already ran.