Implement desired approach to error handling of individual migration/rec failures #34
Description
Context / Goal
We will need to make decisions about how we want failures to be handled and recorded/reported, and which things should be retried.
Expected Outcome
Answer and implement:
- what should happen if one of the two dataset queries fail or are invalid?
- what if an individual row fails for some reason? Should the run be abandoned?
- how should transient errors be addressed? Should the code be able to retry a query in certain situations (network problems etc)
- how should the failures be reported/recorded? Details in logs? Summary error(s) stored alongside the run results with a status?
- how should we distinguish between "user" errors (e.g bad configuration of dataset query) vs "system" errors (unexpected bug in the tool/framework)
- should any (possibly partial) results be removed from the recce DB on failure?
Out of Scope
Additional context / implementation notes