Until now, we have discussed the performance of Replicat because this generally requires tuning, whether it be adjusting the integrated parameters or adding additional processes. Occasionally, however, the Extract process may be slow, but before we tune, we must diagnose. Typically, slow performance lies with I/O operations, but let's perform the following steps to find the root cause for our Extract running in the classic capture mode:
STATS
command:GGSCI (db12server01) 1> stats EOLTP01, totalsonly *, reportrate sec GGSCI (db12server01) 2> stats EOLTP01, totalsonly *, reportrate min
$ cp ./dirprm/EOLTP01.prm ./dirprm/ETEST.prm
TESTMAPPINGSPEED
parameter to the test Extract parameter file, as shown in the following code:TESTMAPPINGSPEED REPORTCOUNT EVERY 5000 RECORDS
ETEST
with begin now
timestamp with the following command:GGSCI (db12server01) 3> add extract ETEST, tranlog, begin now
EXTSEQNO
and EXTRBA
for the archived log, as shown here:GGSCI (db12server01) 4> alter extract ETEST, extseqno <arch_seq_no>, extrba 0
The current list of database archived logs with thread and sequence numbers may be obtained from the following SQL command
SQL> select name, thread#, sequence# from v$archived_log;
GGSCI (db12server01) 5> stats EOLTP01, totalsonly *, reportrate sec GGSCI (db12server01) 6> stats EOLTP01, totalsonly *, reportrate min
TESTMAPPINGSPEED
, then the bottleneck could be the Extract writing to the trail file.TESTMAPPINGSPEED
from the test Extract parameter file and add the TRACE2
parameter in its place to determine whether the Extract is performing any fetch operation.SELECT
statements, and see how long they take to execute.avg
and max
wait times against a given SELECT
statement, the source database has a performance problem; a lack of fresh table statistics or a missing index could be the cause.3.17.23.130