Hello,
I am a sap basis guy, we have recently performed a sap migration :heterogeneous copy from SAP ECC6 Oracle 10.2 Windows to SAP ECC6 Oracle 11.2 HPUX. The whole migration went fine. The ORacle statistics are. of course, up to date
We have just been informed of performance issues when reading the pooled table A017, it turns out that after the migration , the execution plan has changed the primary index the table pool KAPOL is no longer used, and instead it is doing a full table scan leading to dreadful performance :
The ABAP coding generating those full table scan is the following :
select matnr lifnr
knumh
into table lt_a018_1
from a017
for all entries in lt_a018_k
where kappl = 'M'
and kschl = 'PB00'
and lifnr = lt_a018_k-lifnr
and matnr = lt_a018_k-matnr
and ekorg = l_ekorg "'2075'
and werks = l_werks
and esokz = '0'
and datbi >= sy-datum
and datab <= sy-datum.
sort lt_a018_1 by matnr lifnr.
describe table lt_a018_1 lines l_i2.
![13-10-2014 08-31-00.png]()
The Oracle stats are up to date :
![13-10-2014 11-04-55.png]()
As you can see, all the key fields are defined in the where clause, so why the index is no longer used ? That is a total mystery
I did not expect the change of Oracle version to have any effect on the execution plan ...
I understand that similar topic have been discussed in this section of the SCN forum :
http://scn.sap.com/thread/229068
http://scn.sap.com/thread/950128
Sp, maybe some of you (ABAPers) would have some idea, do you understand why the KAPOL index is no longer used ?
Is there anything missing in the ABAP request ?
Any idea would be highly appreciated.
Thank you.