Comments on: Moving Oracle Datafiles to a ZFS Filesystem with the Correct Recordsize

By: Don Seiler

Don Seiler — Thu, 03 Dec 2009 21:48:03 +0000

Michael, you can do it with UNDO. Simply create a new UNDO tablespace in the new ZFS mount. Then change the system undo setting. Wait for all operations to use the new UNDO and you’ll be able to drop the old one.

This is precisely what I did, same goes for TEMP.

By: Michael

Michael — Thu, 03 Dec 2009 21:08:31 +0000

What you are suggesting here will not work for system level tablespaces such as SYSTEM, SYSAUX, UNDO. These will require the entire database to be down to move. They cannot be taken offline while the database is running. You’d still have a database outage!

Also, taking individual datafiles offline in a running system can cause logical data problems which might not be immediately recognizable in an high-throughput application with referential integrity. For example, you could generate an order record in one tablespace, and then, when attempting to assign the order to a customer in an offline tables, you could end up with an orphaned order.

We are encountering this same issue right now. Luckily, the system is not live yet. Good articles such as yours led us to decide to change the ZFS recordsize before we went live (thank goodness)!!!

By: Moving Oracle Datafiles to a ZFS Filesystem with the Correct Recordsize « die Seilerwerks

Sat, 24 Oct 2009 16:20:16 +0000

[...] leave a comment » Originally posted on The Pythian Group blog. [...]

By: Blogroll Report 14/08/2009 – 21/08/2009 « Coskan’s Approach to Oracle

Blogroll Report 14/08/2009 – 21/08/2009 « Coskan’s Approach to Oracle — Mon, 07 Sep 2009 16:41:56 +0000

[...] 10 – Using backup copy to move files after changing the recordsize on filesystem ? Don Seiler-Moving Oracle Datafiles to a ZFS Filesystem with the Correct Recordsize [...]

By: Don Seiler

Don Seiler — Thu, 20 Aug 2009 14:56:38 +0000

Peter: That’s why you have to run the “recover datafile X” command after the switch and before putting the new copy online.

By: Peter Moore

Peter Moore — Thu, 20 Aug 2009 14:38:03 +0000

Hello.

Great article with some interesting ideas.

One question though: Is there not the possibility that the datafile is being written to while the “backup as copy” is running (and until the point when the file is taken offline)?

How do we ensure that the *new* copy of the file is up-to-date?