Statistics Problem ?

Question

Hi,
I have an insert statment that is waiting most of the time on P.K index.
82661 was inserted , but explain plan bellow shows that only : 6396 rows was fetched.
1. Does i have statistics problem ?
2. What would you suggest to do in case of waiting on db file sequential read (please remember that its a primark key index) .
(remarks: I cant use NOLOGGOING , i set initrans to 20 )
Thanks
INSERT /*+ APPEND */ INTO TLOG (USERNAME,ACCT_SESSION,IP,START_DATE,END_DATE,
CALLING_NUMBER,CALLED_NUMBER,NAS_IP,
NAS_TYPE,NAS_PORT,START_STOP,SERVEREND_POINT,CLIENTEND_POINT,SERVER_AUTH_ID,CLIENT_AUTH_ID,
RB_CONTENT_NAME,ACCT_INPUT_OCTETS,ACCT_OUTPUT_OCTETS,SPEED,TERMINATE_CAUSE,ACCT_SESSION_TIME,
EVENT_TIMESTAMP,ACCT_AUTHENTIC,ACCT_DELAY_TIME,ACCT_INPUT_PACKETS,ACCT_OUTPUT_PACKETS,
FRAMED_PROTOCOL,SERVICE_TYPE,TERM_CODE)
SELECT /*+ PARALLEL (p,8) */ USERNAME,ACCT_SESSION,IP,START_DATE,END_DATE,CALLING_NUMBER,
CALLED_NUMBER,NAS_IP, NAS_TYPE,NAS_PORT,START_STOP,SERVEREND_POINT,CLIENTEND_POINT,
SERVER_AUTH_ID,CLIENT_AUTH_ID,RB_CONTENT_NAME, ACCT_INPUT_OCTETS,
ACCT_OUTPUT_OCTETS,SPEED,TERMINATE_CAUSE,ACCT_SESSION_TIME,EVENT_TIMESTAMP,ACCT_AUTHENTIC,
ACCT_DELAY_TIME,ACCT_INPUT_PACKETS,ACCT_OUTPUT_PACKETS,FRAMED_PROTOCOL,SERVICE_TYPE,TERM_CODE
FROM RADIUS_LOG_EXT p
LOG ERRORS INTO ERR_LOG REJECT LIMIT unlimited
call     count       cpu    elapsed       disk      query    current        rows
Parse       10      0.24       0.23          0       7809          0           0
Execute     10     25.17     288.97      47830      37301    1021273       82661
Fetch        0      0.00       0.00          0          0          0           0
total       20     25.41     289.21      47830      45110    1021273       82661
Misses in library cache during parse: 9
Optimizer mode: ALL_ROWS
Parsing user id: 59  (RADIUS)
Rows     Row Source Operation
   6396  PX COORDINATOR  (cr=36 pr=1 pw=0 time=319409 us)
      0   PX SEND QC (RANDOM) :TQ10000 (cr=0 pr=0 pw=0 time=0 us)
      0    PX BLOCK ITERATOR (cr=0 pr=0 pw=0 time=0 us)
      0     EXTERNAL TABLE ACCESS FULL LOG_EXT (cr=0 pr=0 pw=0 time=0 us)
      0  INSERT STATEMENT   MODE: ALL_ROWS
   6396   PX COORDINATOR
      0    PX SEND (QC (RANDOM)) OF ':TQ10000' [:Q1000]
      0     PX BLOCK (ITERATOR) [:Q1000]
      0      EXTERNAL TABLE ACCESS (FULL) OF 'LOG_EXT' (TABLE)
                 [:Q1000]
Elapsed times include waiting on following events:
  Event waited on                   Times   Max. Wait  Total Waited
  ------------------------------   Waited  ----------  ------------
  SQL*Net message to client            11        0.00          0.00
  SQL*Net message from client          11        0.00          0.00
  os thread startup                     8        0.01          0.13
  PX Deq: Join ACK                     52        0.00          0.00
  PX Deq: Parse Reply                  64        0.04          0.35
  PX Deq: Execute Reply                56        0.01          0.10
  db file sequential read           47824        0.23        265.09
  buffer busy waits                     8        0.00          0.00
  undo segment extension                4        0.00          0.00
  PX Deq: Signal ACK                   92        0.00          0.00
  enq: PS - contention                 25        0.00          0.00
  log file sync                        11        0.04          0.14
**********************************************************************

Answer 1

user10634138 wrote:
Hi ,
I remove the parallel hint from the select statment but i am still facing poor insrert performance.
Now its took 373 seconds (5.5 minute) to insert just : 56,483 rows as you can see bellow.
It's been a few years since I last checked, but the LOG ERRORS may be turning your array-based insert into single row processing, leading to excess random I/O.
You may find that you can load the data faster if you use a pl/sql approch that does a "bulk collect and forall insert" with array exeception handling, processing (say) 100 rows at a time. You may get an extra benefit if your driving select also does an 'order by' on the primary key.
Do you have only one index on the table ? The number of current gets seems a bit too high for the number of rows inserted if there's only one index.
Regards
Jonathan Lewis
http://jonathanlewis.wordpress.com
http://www.jlcomp.demon.co.uk
To post code, statspack/AWR report, execution plans or trace files, start and end the section with the tag {noformat}{noformat} (lowercase, curly brackets, no spaces) so that the text appears in fixed format.
"Science is more than a body of knowledge; it is a way of thinking"
Carl Sagan