We are currently looking into the recommendation of having a single extract / replicat replication group.
I may change my opinion on recommending the number of extracts / replicats for a particular customer, I have based this on some information in discussions some really knowledgeable folks. I am going to consider using this as a guideline from now on to create a single extract details to follow.....
Best Practices for HA Off-Box and Hub Configs
OGG_bp_instantiation_Oracle_DocID_1276058.1
This was received from Mike Papio from the A-Team for import methodology for GoldenGate
Oracle 19c Goldengate downloads
There are several types of supplemental logging
Minimal supplemental logging ensures that products leveraging LogMiner technology will have sufficient information to support chained rows and cluster tables.
Primary key supplemental logging includes the primary key for rows affected by UPDATE and DELETE changes.
Unique key supplemental logging includes all columns for a unique key are written to undo if any unique key columns are modified.
Foreign key supplemental logging includes all other columns belonging to a foreign key will be logged in the undo if any foreign key columns are modified.
If no primary key or unique key is available then it is possible to specify that all columns are logged. In this case all columns in a row will be logged in the undo. When the row is replicated in the target database, equality predicates will be applied to all columns. LONG, LONG RAW and LOB columns will be omitted from the supplemental logging.
Procedural replication supplemental logging includes additional information in the redo log during invocation of procedures in Oracle-supplied packages for which procedural replication is supported. I have never investigated this option.
Oracle implements supplemental logging at database level, schema level and at table level. The three implementations are significantly different:
Database level supplemental logging is configured in the control file. The parameter does not appear to be stored in the database itself.
We were doing some testing for GoldenGate and decided to put a little article together for doing the export / import
3 things
Do not import indexes
Do not allow creation of stats on import
Use parallelism (2x cpu_count)
logdump
ghdr on
detail on
open <trailfile name>
count
This will detail the record count
Jinyu's blog is a great description of how to handle these