Why can we have UPI only when PPI column is part of PI, can someone explain in detail?
Thanks in Advance
If the column was not part of the PPI TD should check all and every partition to guarantee unicity, which has a high cost.
That's why. (it's in the manuals, BTW)
If you really want to check uniqueness in this case, then you can create a USI for the NUPI column(s) for the partitioned table. This would use the USI index to reduce the high cost which Carlos has explained.