How to set up Replicat Process with GoldenGate on Exadata & DBFS?

Question

Hi all,
I confuse with usage of DBFS on Exadata when trying to config Replicat Process via this Oracle document:
[Oracle Best Practices For High Availability|http://www.oracle.com/technetwork/database/features/availability/maa-wp-gg-oracledbm-128760.pdf]
I have a system with these description:
EXADATA: 2 node ExaDB01 & ExaDB02. This system mount a DBFS /mnt/dbfs/ggs for store GoldenGate subdirectory (trail and checkpoint file only)
FUJITSU: 2 node FujDB01 & FujitDB02.
I need to config GoldenGate with DataPump to sync data between these two systems.
On ExaDB01, i created an Extract Process to extract trail file to EXTTRAIL location /mnt/dbfs/ggs/dirdat/aa.
Now, I need to config datapump on ExaDB01 and Replicat Process on FujDB01. At this point, i confuse with DBFS configuration for EXTTRAIL & RMTTRAIL:
1) At ExaDB01:
GGSCI (exadb01) 1> ADD EXTRACT exadump, EXTTRAILSOURCE /mnt/dbfs/ggs/dirdat/aa
GGSCI (exadb01) 1> ADD RMTTRAIL /mnt/dbfs/goldengate/dirdat/aa EXTRACT exadump, MEGABYTES 500
2) At FujDB01: Replicat Process
GGSCI (fujdb01) 1> ADD REPLICAT rexadb, EXTTRAIL /mnt/dbfs/goldengate/dirdat/aa
So in this case:
Q1) ExaDB01 and FujDB01 have the same mount point /mnt/dbfs/ggs?
Q2) With ExaDB01: The "local trail" and "remote trail" in this case are stored in the same directory /mnt/dbfs/ggs/dirdat?
Q3) If "local trail" and "remote trail" are stored in the same directory. So why we need to config DataPump to push "local trail" to "remote trail" location?
Please light me up!
Tks & Brgd
Edited by: Dung Nguyen on Sep 7, 2011 11:10 AM
Edited by: Dung Nguyen on Sep 7, 2011 11:12 AM

Answer 1

GoldenGate forum is over here:
GoldenGate
(thread has moved from Replication to GoldenGate)
Extract
EXTTRAILl aa is created
Data Pump (secondary extract)
Reads and sends your exttrail to a remote server, which can also be the same server, so in that case, you need to rename the file (the trail ID, two character prefix).
The linkage is you add the data pump extract, and then add the rmttrail
ADD EXTRACT <dp name>, EXTTRAILSOURCE wherever/aa, BEGIN <time>
ADD RMTTRAIL <wherever on your server/bb, EXTRACT <same dp name>
Note that the trail ID has switched from aa to bb. If going to a remote server, you can use the same ID. If going to the same server, put it in a different path, or have the trail read by replicat named something else via the ID.
If you wanted to use the same dirdat directory, then your original extract is created with aa, and your trails read by replicat can read bb - after the pump has "sent" the aa files over as bb.
You don't need a data pump, but as a best practice, always use one. Using a data pump eliminates replication failure due to network outage. If not using one, then what happens when extract has nowhere to send a trail to on a remote server (because it cannot reach/communicate with the remote server)?
If using the same server without a data pump, then your primary extract "sends" or creates the trail in the rmttrail location, where replicat then reads from. In that case, you only have the one trail ID.
If one on server, you can also create a second instance of GoldenGate (in its own home), so it will have a slightly different path, and in that case, you do not have to switch the trail ID because the path difference will take care of separating the files.
Edited by: stevencallan on Sep 7, 2011 1:13 PM