We currently scan all column families in the destination row when performing a foreign key check. We should limit this to the primary column family, which we know will never be missing as long as the row exists.
This has implications on transaction contention. For instance, see #30624.
cc. @knz for triage.
We currently scan all column families in the destination row when performing a foreign key check. We should limit this to the primary column family, which we know will never be missing as long as the row exists.
This has implications on transaction contention. For instance, see #30624.
cc. @knz for triage.