Two Thoughts
Patrick Jolliffe, August 18, 2018 - 6:42 am UTC
Reading this response, I had two thoughts.
Firstly, I think 10053 trace might be a bit drastic at this stage, sure it contains lots of interesting information, but the volume of information can be a bit overwhelming especially if one is not reasonably knowledgeable about the Cost-Based Optimizer. I'd start by picking the most simple problem query and (assuming appropriately licensed) using Real-Time SQL Monitor, comparing estimated with actual cardinalities to try to find out where things are going wrong. If you don't have license for that, hinting with GATHER_PLAN_STATISTICS and using dbms_xplan.display_cursor with ALLSTATS LAST can be used similarly (google it). You may have to drill into 10053 trace eventually, but keep it simple first.
Secondly, I guess you are where you are, but in 2018 I can't understand why anyone would be going from 11.2 to *only* 12.1. 12.1 is already out of Premier Support (Extended Support fee waived until July 2019). If you're going to have to deal with the pain of the upgrade at least go to 12.2 (if you think 18c is a bit too new for you). Not only do you get huge amount of issues with 12.1 resolved properly (adaptive statistics spring to mind), you are then within Premier Support until 2022.
August 18, 2018 - 12:36 pm UTC
Nice input.
I'm leaning toward 10053 as a first step, not to tune the SQL but to see if a pattern can be found, which can then perhaps lead to a global solution, eg, "Ah... the system stats are making multiblock reads too cheap"
Because if a cause cannot be found, then we're into bad territory...because this is a SAP system - you can't go tinkering with the SQL's.
What About upgrade itself
Gh, August 18, 2018 - 6:55 am UTC
You didn't talk about upgrade itself. How it was completed and what about os characteristics?
For instance did you gathered system stats?