Hi all,
We have a requirement to migrate seven data warehouse databases from 11.2.0.4 on AIX (so Big Endian) to 19c on RHEL7 (Little Endian). Total size across the seven databases is ~12TB and, Golden Gate having been ruled out by our management, we're currently looking at using FTEX.
These databases are pretty old and each one has ~30 tablespaces (for small/medium/large tables/indexes) which, as part of the cross-platform migration, we would like to rationalize into three tablespaces - for tables, indexes, and LOBS. Not only does this sit better with norms which have been put in place since the databases were created, but it is also likely to make the migration process less painful.
The question I have is related to tablespace block sizes. Some source databases have both 8K and 32K tablespaces (although some of these only have an 8K buffer cache!) while others have just 32K tablespaces, (again with an 8K cache!) and the logic of 'large objects in a 32K tablespace' and 'smaller objects in an 8K tablespace' isn't applied uniformly either.
In addition the 'standard' block size, used by the ~100 databases we've already migrated to 19c on RHEL7 is 8K, so there's an argument (however flawed) for using 8K for our seven data warehouses.
Any thoughts so far would be much appreciated. Anything I'm not taking into account which I should be ? Any flawed decision making ?
I arrived here because I landed on Domagoj's post - https://community.oracle.com/tech/developers/discussion/541465/moving-table-to-a-tablespace-with-different-block-size and, while the responses are clear and make sense (esp. Jonathan's !!), I'm starting to wonder whether it's worth the effort or whether I can achieve my goal differently.
Thanks, 'O wise ones of the Oracle Groundbreakers Developer Community', for any time/neurones you're willing to spend on this thread and any observations you're able to share.
Regards,
Jes