Re-run your test (if necessary) then
select * from table(dbms_xplan.display_cursor('{the sql id from the trace
file}',null,format=>'outline'));
You've got a FILTER operation at the start of the plan, and it looks like
it's a "conditional SQL"*** filter that results in Oracle not executing the
rest of the plan.
The obvious guess for this is that you've got a security predicate being
generated for this query which ensures that only the correct users can see
the data. Using dbms_xplan.display_cursor we'll see all the predicates, and
the predicate on the FILTER operation may be something as simple as "1=0".
*** See https://jonathanlewis.wordpress.com/2007/01/09/conditional-sql/
Regards
Jonathan Lewis
On Fri, 13 Aug 2021 at 04:17, Hameed, Amir <amir.hameed@xxxxxxxxxxxxxxx>
wrote:
Hi,
The database is a three-node 11.2.0.4 RAC and the OS is Solaris 11. This
is an Oracle EBS applications environment and the code in question is
standard Oracle code. We ran a concurrent job with tracing enabled as part
of troubleshooting performance issues with that job. The query in question
looks pretty bad because of the way it ran:
SELECT DECODE(NVL(CCH.TRANSACTION_AMOUNT, 0),0, CCL.COMMISSION_RATE,
ROUND(CCL.COMMISSION_AMOUNT/CCH.TRANSACTION_AMOUNT, 7))
FROM
CN_COMMISSION_HEADERS CCH, CN_COMMISSION_LINES CCL WHERE
CCH.COMMISSION_HEADER_ID = CCL.COMMISSION_HEADER_ID AND
CCH.SOURCE_TRX_ID =
:B12 AND CCH.SOURCE_TRX_LINE_ID = :B11 AND CCL.CREDITED_SALESREP_ID =
:B10
AND CCL.QUOTA_ID IN (:B9 ,:B8 ,:B7 ,:B6 ,:B5 ,:B4 ,:B3 ,:B2 ) AND
CCL.STATUS = :B1
call count cpu elapsed disk query
current rows
------- ------ -------- ---------- ---------- ---------- ----------
----------
Parse 11 0.00 0.00 0 0
0 0
Execute 3832 0.13 0.13 0 0
0 0
Fetch 3832 78354.13 137942.64 2415515000 2418912980
0 3793
------- ------ -------- ---------- ---------- ---------- ----------
----------
total 7675 78354.27 137942.78 2415515000 2418912980
0 3793
Rows (1st) Rows (avg) Rows (max) Row Source Operation
---------- ---------- ----------
---------------------------------------------------
1 1 1 NESTED LOOPS (cr=631072 pr=630185 pw=0
time=704520 us cost=179182 size=58 card=1)
3 2 10 NESTED LOOPS (cr=630896 pr=630016 pw=0
time=125749 us cost=179182 size=58 card=2)
1 1 5 TABLE ACCESS FULL
CN_COMMISSION_HEADERS_ALL (cr=630892 pr=630015 pw=0 time=618482 us
cost=179177 size=26 card=1)
3 2 10 INDEX RANGE SCAN
CN_COMMISSION_LINES_N10 (cr=4 pr=0 pw=0 time=4556 us cost=3 size=0
card=2)(object id 384194)
1 1 1 TABLE ACCESS BY INDEX ROWID
CN_COMMISSION_LINES_ALL (cr=7 pr=1 pw=0 time=9561 us cost=5 size=32 card=1)
Elapsed times include waiting on following events:
Event waited on Times Max. Wait Total
Waited
---------------------------------------- Waited ----------
------------
library cache lock 1 0.00
0.00
library cache pin 1 0.00
0.00
db file sequential read 6136 0.08
52.33
enq: KO - fast object checkpoint 9900 0.31
15.96
reliable message 3832 0.55
12.97
direct path read 93230079 0.12
61862.37
Disk file operations I/O 776 0.04
1.03
latch: cache buffers chains 1 0.00
0.00
read by other session 173 0.01
1.16
gc cr grant 2-way 1836 0.00
0.73
KJC: Wait for msg sends to complete 43 0.00
0.00
latch free 14 0.00
0.00
gc buffer busy acquire 2 0.00
0.00
kfk: async disk IO 1 0.00
0.00
gc current grant 2-way 1 0.00
0.00
********************************************************************************
I pulled variables of this statement from V$SQL_BIND_CAPTURE, defined them
in a script, and manually ran the statement with event 10046 at level 8:
SELECT DECODE(NVL(CCH.TRANSACTION_AMOUNT, 0),0, CCL.COMMISSION_RATE,
ROUND(CCL.COMMISSION_AMOUNT/CCH.TRANSACTION_AMOUNT, 7))
FROM
CN_COMMISSION_HEADERS CCH, CN_COMMISSION_LINES CCL WHERE
CCH.COMMISSION_HEADER_ID = CCL.COMMISSION_HEADER_ID AND
CCH.SOURCE_TRX_ID =
5920213 AND CCH.SOURCE_TRX_LINE_ID = 33025874 AND
CCL.CREDITED_SALESREP_ID = 100089722
AND CCL.QUOTA_ID IN (1441 ,1587 ,1461 ,1588 ,1481 ,1592 ,1201 ,1593 ) AND
CCL.STATUS = 'CALC'
call count cpu elapsed disk query
current rows
------- ------ -------- ---------- ---------- ---------- ----------
----------
Parse 1 0.14 0.15 0 0
0 0
Execute 1 0.00 0.00 0 0
0 0
Fetch 1 0.00 0.00 0 0
0 0
------- ------ -------- ---------- ---------- ---------- ----------
----------
total 3 0.14 0.15 0 0
0 0
Rows (1st) Rows (avg) Rows (max) Row Source Operation
---------- ---------- ----------
---------------------------------------------------
0 0 0 FILTER (cr=0 pr=0 pw=0 time=1 us)
0 0 0 NESTED LOOPS (cr=0 pr=0 pw=0 time=0 us
cost=179182 size=52 card=1)
0 0 0 NESTED LOOPS (cr=0 pr=0 pw=0 time=0
us cost=179182 size=52 card=2)
0 0 0 TABLE ACCESS FULL
CN_COMMISSION_HEADERS_ALL (cr=0 pr=0 pw=0 time=0 us cost=179177 size=23
card=1)
0 0 0 INDEX RANGE SCAN
CN_COMMISSION_LINES_N10 (cr=0 pr=0 pw=0 time=0 us cost=3 size=0
card=2)(object id 384194)
0 0 0 TABLE ACCESS BY INDEX ROWID
CN_COMMISSION_LINES_ALL (cr=0 pr=0 pw=0 time=0 us cost=5 size=29 card=1)
The statement ran very quickly. The two plans look the same. The one
difference is that in the second plan, FTS on table
CN_COMMISSION_HEADERS_ALL performed zero physical reads (pr=0) which is
baffling because the table has over 10M rows. In the first plan, FTS was
done using DIRECT PATH READS and therefore, the data couldn’t possibly have
been cached in the Buffer Cache. I flushed the Buffer Cache just to see if
that would make any difference but it didn’t, as expected.
I am trying to understand when the statement was run manually and it did
do FTS of the table, why there were no physical reads reported.
Thanks,
Amir
