You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I think the problem is that each rewrite task calls reserve_fragment_ids itself, which can cause contention on the commit if there are a lot of threads. Each task needs to do this so that it can construct the row id map. (We don't have accurate target row ids until the fragment ids are established.)
The stable row ids will be able to move the reserve fragments to the end, removing this contention.
For existing row ids, maybe we can do something more clever to make this possible.
repro
The text was updated successfully, but these errors were encountered: