Transaction block consistent read are high after upgrade to 12c database
we upgraded our database from 11gr2 to 12cR1 last year. As soon as after the upgrade. Our Queries started to run longer. And it was never ending. The plan and wait events looked similar.
So what changed? We did our analysis and found out that the queries are reading data from undo. And found it to be delayed block cleanout with stats type - Transaction block consistent read- undo records applied.
We the set the _rollback_segment=2000 AUM. After setting the value the undo issue somehow reduced. but not completely. Something has changed in 12c undo management feature which is sadly effecting us badly.