GG on Teradata

784624

Hi

We are evaluating Golden gate to replicate data from Oracle Source database to Teradata on Unix. I have few questions and need your expert comments.

1) How can we do the initial Load on Teradata - We have huge tables(>500GB) on the source database.I guess the extract specialrun may not be efficient.Do we have to rely on Teradata tools for the initial Load.

2)GG uses ODBC to insert records into Teradata .Can someone point to some documents which would help us to configure ODBC on Unix/Linux for Teradata target.

Appreicate if you can share your experience on Teradata with GoldenGate.

Thanks

-joe

user10291122,


1) You can use the Teradata tools or use OGG initial loads. The OGG "direct" load (see the OGG general Admin and Reference guids) bypasses disk and is relatively fast but with delays in reading from the source and writing to the target you may blowout your undo/rollback. In addition, if the target fails (e.g. running out of space on the first initial loads happens a lot) you have to redo that run. Therefore I suggest using something that isn't documented but is not secret: write the initial load data to a proper set of trails files. This means you will then setup a "normal" replicat (ideally several) to load the data and possible use a pump to pump the data to the target if your initial extract writes local, which is recommended because this will be the fastest way to unload the data.

Here's the trick:

ggsci> add extract e_initaa, sourceIsTable

Paramfile:
extract e_initaa
userid ggs, password ggs
extFile ./dirdat/aa, maxFiles 999999, megabytes 500
Table scott.*;

Notice the usage of "maxfiles" here. Normally you'll use several extracts, perhaps one per table or you can even use a WHERE clause in the table statement to sub set a table across many extracts. A best practice here is to use a clone of the source, recover it to an SCN (say, 123456) then do your loads from that temporary instance or just borrow some time on a backup instance. Once your initialization is done you will start your real-time replicat (not the intial load replicat) using the AFTERCSN feature, as in:

ggsci> start repaadb aftercsn 123456

2) We're working on posting the Teradata installation guide but for now just open an SR and request it and they will send you a copy.

Good luck,
-joe

784624

Joe

Thank you for your response. We have to process Inserted records in the source database only. Is the formatascii extract trail format compatible with any teradata utility(Loaders). This way we can avoid configuring replicat process on teradata end.

-joe

I found the OGG for Teradata on eDelivery already there but a little tucked away.

For "Oracle GoldenGate Documentation on Non-Oracle Databases" (product V21825-01) go here:
http://edelivery.oracle.com/EPD/Download/get_form?egroup_aru_number=12055596

Why not just try a replicat first and see what throughput you get before you assume it's slower than the Teradata load tools? You can start loading data into the target in parallel and without waiting until it's all unloaded form the source.

Good luck,
-joe