I will need replicate some tables (in uni-direcional mode only, ie,
from many sources to the only dest db) and refreshing data nightly) ,
sources being some medium-size tables (a few Gbs in size) in SQLServer
2000, Oracle 9ir2EE and Oracle 10gr2 EE databases (in diverse servers,
some running Windows, some linux, but all of them network-accesible)
to a 10gr2 EE destination db : regarding the Oracle source dbs, I
tested replication with MVs with success, but I´m considering CDC too,
but I have some doubts to be smashed from the real experiences from
the members :
a. some Oracle source dbs are in noarchive mode : in the "OracleŽ
Database Data Warehousing Guide 10g Release 2" manual, chapter 16 -
Change Data CaptureAsynchronous Distributed HotLog Mode , it says :
"In the asynchronous Distributed HotLog mode, change data is captured
from the online redo log file on the source database"
So as I understand the noarchive mode in the source dbs are not a
problem, right ? I question it because the entry in
http://www.rittmanmead.com/2006/04/1...and-owb-paris/ puts archivemode
as a need , togheter with activated global_name parameter, as some
metalink notes do, is it right or wrong ?
b. some LONGs and LONG RAWs are present, unfortunately, and can´t be
changed right now, and all the docs says that LONGs are not supported
in replication of any type : can I create a view "translating" the
LONGs to LOBs and use the view as the data source ? If not, some small
example of triggers doing it ? Or other option to replicat this
things ?
c) speaking in triggers, if I switch to synchronous CDC (the made with
triggers one), is possible to refresh only at certain times ? The
quoted manual says that sync CDC is in immediate, commit-base only,
BUT imho if it maintains a log table (and it does) , maybe could be
possible another way of refreshing...
d) any gotchas, recommendations and alike for this scene ?
Regards,
Chiappa