Details
-
Improvement
-
Status: Resolved
-
Low
-
Resolution: Fixed
-
None
-
None
Description
it's easy to handle skipping bad rows during compation in the PreCompacted (merged-in-memory) path and we have done this for a long time. It is harder in the LazilyCompacted path since we have already started writing data when we discover that some of the source rows cannot be deserialized. This adds mark/reset to SSTableWriter so compaction can skip back to the beginning in these circumstances.