3 Replies Latest reply on Dec 28, 2012 3:50 PM by stevencallan

    Initial Load using expdp/impdp

      Golden Gate version:
      RDBMS version :
      OS : RHEL 5.4

      From Production OLTP (Source) to Production ETL (Target), we want to replicate 5 tables.
      Each table is around 3 gb in size in Source now.
      In Source , around 200 mb of data is being INSERTed every hour in all these 5 tables combined .

      This is how we are planning to do the Iniitial Load:
      After installing goldengate on both servers (and keeping the parameter files ready), we are going to take and exdp dump of these 5 tables from Source and SCP it to target server and import it there using impdp.

      Total estimated time for expdp, scp the dumpfile and impdp in target DB is: 2 hours

      During this gap of 2 hours , 400 mb worth of data would have been INSERTed in these 5 tables in the source. We want the tables to be in Sync in both Source and Target. For that ,
      1. Shouldn't we be leaving the extract at the source running so that it will capture the changes during these 2 hours ?
      2. What about the secondary extract (Pump) in source. Should it be running ?
      3. Replicat in the Target should be stopped during this period of 2 hours. Right ?
                          Anything else we have to do to make this configuration smooth ?
        • 1. Re: Initial Load using expdp/impdp

          2-Not necesarily, although once you want to start replicat on the target you should have the files available there. If the replicat on target is stopped already it should not make a difference if this is stopped or not
          3-Correct, else it will most liekly fail with row not found when trying to apply a transaction.

          I assume you will be using flashback_scn method?

          Configure and start the extract processes in the source side
          Take an export of that schema using flashback_scn number
          Import into the target database
          Configure the replicats
          Start the replicats after SCN# like start replicat <rep_name>, aftercsn <SCN#>

          N K
          • 2. Re: Initial Load using expdp/impdp
            Thank you NK for reminding us about FLASHBACK_SCN . We would've missed that and it would have caused all those FK errors.

            You mentioned about starting replicat with aftercsn clause.
            start replicat <rep_name>, aftercsn <SCN#>
            You mean I have to get the SCN of the source using either of the below queries just before I start the expdp of the source tables ?
            select current_scn from v$database;
            select dbms_flashback.get_system_change_number() from dual;
            • 3. Re: Initial Load using expdp/impdp
              Just follow the setup steps shown in the tutorial: