I would like to have your ideas regarding how to optimize backup/restore in a VLDB Oracle Database
- RAC 19c 4 Nodes
- Version 19c
- Size: 32TB
- Mode: NOARCHIVELOG
Let me explain some considerations regarding why the database is in NOARCHIVELOG mode:
- There is no single process which writes data into the database in
online mode. All the processes are batch and the data is always in
external systems that can redeliver the data in case it is needed.
- All the tables are in NOLOGGING mode.
- The Service Level Agreement ( SLA ) for some of the processes is really critical, and we did
tests with the database in ARCHIVELOG and the results were not good enough to cover such SLA.
My current strategy is as follows:
- Every First Sunday I run an INCREMENTAL LEVEL 0
- Every Sunday ( except the first one ) I run an INCREMENTAL LEVEL 1 CUMULATIVE
- I have a backup window of 30 hours
- The backup is done using AS COMPRESSED BACKUPSET to a Tape drive in TSM using the four nodes of the RAC ( 16 channels in total )
- Historic Tablespaces are in READ ONLY and the backup uses SKIP READ ONLY.
- I use filesperset and section size to improve performance.
I am not worry about the backup window, the level 0 is taking 17 hours, and the level 1 is taking between 2 and 4 hours ( the last cumulative before the next level 0 ).
The only thing that worries me is if I lost ASM and I need to restore/recover from the backup. I have been reading about RMAN ROLLING FORWARD INCREMENTAL using COPY, but I am afraid the copy will take ages when the destination is TAPE.
Do you believe there could be a better strategy than this one for restore/recover purposes?
Many thanks in advance