Hello,
We are running SAP ECC6 EhP5, Oracle 11.2.0.2, HP-UX 11.31 ia64
UNO_RETENTION = 21600
The size of the tablespace PSAPUNDO is fixed (AUTOEXTENT=NO)
We have been running Oracle Undo Management features for year, without any problem ... until a few weeks ago.
We noticed a problem at the SAP level, with the sm12 locks not being released, it was a problem with the SAP enqueue server, and we have set the SAP parameter :
enque/deque_wait_answer = TRUE Wait for answer from enqueue server (synchronous)
We haven't done any other change, and certainly not changed any Oracle paramter
It solved our SAP sm12 problem , but now, after the ORacle/SAP restart we are observing a change in the ORacle Undo behaviour :
First :
After the restart, the tablespace PSAPUNDO is now constantly full, and extending the tablespace is of no effect.
I undersatnd that PSAPUNDO being full is not necesarily a problem, but in our case it is always full, we never had this before and the user activity has not increased
We not frequently have severa loracle errors related to that tablespace : ora-1515,ora 30036, ora 1652 : jobs, dialog and update activities are cancelled
Second the Oracle views related to undo management do not work as they used to :
As you can see in the screenshot, the dba historical view of Undo management is no longer filled after the system restart.
As designed, there used to be one entry, generated every 10 minutes
There is now, one single entry in V$UNDOSTAT whic hshows a very low value for TUNED_AUTORETENTION : 4628 sec
We have checked ORacle paramters change history, absolutely no changes have been done.
We are really perplex :
What could cause the PSAPUNDO tablespace to be constantly full , all of a sudden , could it be the sap parameter enque/deque_wait_answer ?
What could cause the ORacle view V$UNDOSTAT to only keep one single entry ?
The size of the tablespace PSAPUNDO is fixed (AUTOEXTENT=NO), does that mean that the UNDO_RETENTION value is ignored ?
Thanks and Regards