Today we got a request to refresh a development database from a copy of production. I decided to use the canned clone process that ships with OEM 12c instead of the usual scripts or manual processes. I was pleasantly surprised with the results. There was a failure related to an NFS problem and I was able to restart the process easily.
Pro:
- Minimizes the amount of scratch disk space by backing up and restoring one datafile at a time.
- No scripting or manual process required.
- Uses standard RMAN under the covers.
- Works with older versions (This one was 10.2)
- Works with existing backups or directly from the source database (This database was not previously backed up and was not in archivelog mode)
- Can duplicate a live database if it is in archivelog mode and running at version 11
- Easy to restart after errors.
- Runs as a database job in the background, unattended.
- Works on the same machine or across the network.
Con:
- More time required – one datafile at a time, no parallelization. 763G ran for 6.5 hours.
- No email notification for errors or completion.
- Black box – you really need to know how to do the steps manually and this will not teach you.
- Needs scratch space on local disks to avoid NFS performance problems. (size of largest datafile)