Connect CDC SQData

In an Oracle RAC there may be more than one node or Oracle Instance in operation at any one time. The capture should be configured to connect to a Scan address rather than an individual Node address through TNS. The Scan address connects to the Cluster not an individual Oracle Instance. In the following example the tnsnames.ora configuration file for a two node cluster might  look like this:

# tnsnames.ora Network Configuration File: /opt/app/oracle/product/19.3.0.0/db_1/network/admin/tnsnames.ora

# Generated by Oracle configuration tools.

orcl.customer.com =

 (DESCRIPTION =

   (ADDRESS = (PROTOCOL = TCP)(HOST = ora-rac-19-scan)(PORT = 1521))

   (CONNECT_DATA =

     (SERVER = DEDICATED)

     (SERVICE_NAME = orcl.customer.com)

   )

 )

n1.orcl.customer.com =

 (DESCRIPTION =

   (ADDRESS = (PROTOCOL = TCP)(HOST = ora-rac-19-n1)(PORT = 1521))

   (CONNECT_DATA =

     (SERVER = DEDICATED)

     (SERVICE_NAME = orcl.customer.com)

   )

 )

n2.orcl.customer.com =

 (DESCRIPTION =

   (ADDRESS = (PROTOCOL = TCP)(HOST = ora-rac-19-n2)(PORT = 1521))

   (CONNECT_DATA =

     (SERVER = DEDICATED)

     (SERVICE_NAME = orcl.customer.com)

   )

 )

The connection string in the capture.cab file might then look like this when connecting to the "cluster" rather than one of the two nodes:

$ sqdconf create oracle.cab --type oracle --database=dbname1:1521/orcl.customer.com

--user=uuuuuuu --password='ppppppppppp' --store=/home/sqdata/oracdc1/oracdc1_store.cab

While this example will connect the capture to one of the two Oracle nodes, it it will not automatically fail over should the node we are directed/connected to actually fail. When that happens the connection to the Log miner is lost and and any data that has been mined but not yet published and acknowledged by a downstream Engine will have to be re-mined when the capture reconnects. The capture configuration .cab file maintains the safe restart SCN/LSN's for the capture and the individual subscribing Engine(s). When the Capture reconnects to any other node in the cluster it will then request the list of available logs from Oracle and resume mining from the safe restart point, losing NO data in the process.

The capture does not however automatically reconnect and will require the sqdconf start command to be run. This should not be a regular occurrence in any production environment.