rmpifer commented on pull request #2342: URL: https://github.com/apache/hudi/pull/2342#issuecomment-749217844
> @rmpifer On this point specifically, > > > so there would never be the case where a rollback was being written before an instant earlier on the timeline was already synced. Removed this logic because it created circular dependency when implementing timeline merging > > Following situation is technically possible. Timeline has commits c1, c2, c3, c4 and we have synced only upto C2. User decides to restore the table to commit time c2, thus generating rollback metadata for c3, c4 which now have no files in the metadata table. I think we should defensively handle this scenario at the payload merge level i.e if there is a delete for a file that does not exist in the base file or prior instants, we should just do a no-op and continue. What I am saying is : I agree with you that the extra logic need not be present, but we should check the payload once to ensure we can handle the scenario above. > > Let me know if we are not talking about the same thing :) If this is the case what happens when c3 and c4 are then finally synced? These files would remain in the metadata table even though rollback should have removed them ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected]
