Hello i have a the following case, can you tell me...
# troubleshooting
g
Hello i have a the following case, can you tell me what will happen? We are a RealTime table without upsert that has time and partition pruning, during consumption there is a problem with a single record from application POV (record is skipped and not published), now other records with newer timestamps are published, can we then backfill this lost record with new kafka timestamp but old out of order creation time (used for the time pruning), will the somehow break the pruning process or it will manage to fix itself?
m
Yes you can backfill old records. Time pruning happens based on min and max time within a segment. The older record will just impact the min value (if it is older than current min seen), but the pruning would still apply
g
so i guess in theory it would scan some additional segments
that is really good that it works this way
m
Yes