Thanks for the question, Evgenii.
Asked: January 10, 2020 - 7:55 am UTC
Answered by: Connor McDonald - Last updated: January 13, 2020 - 4:32 am UTC
Category: Database Administration - Version: 12
Viewed 100+ times
I'm goint to start using two 3TB NVMe PCI disk drives - for Oracle 12 ASM DATA Diskgroup.
What can be a best practice for redundancy in this case? -
- to use Linux Software Raid (Mirroring) on 2 disk drives, and run Diskgroup with ASM External Redundancy on top of /dev/md1,
- or give 2 raw disk drives to ASM, and let ASM deal with mirroring with "Normal Redundancy" ?
Setup in more details:
- Oracle Database 12c Enterprise Edition Release 18.104.22.168.0 - 64bit Production
- Red Hat Enterprise Linux Server release 7.4
- ProLiant DL360 Gen10 server,
- 2 pcs HPE 3.2TB NVMe x8 Lanes Mixed Use HHHL MT003200KWSTC Card
and we said...
I think either will be fine, but if those drives are being used for nothing but Oracle, I would typically go with ASM because the greater the control over the stack, the more confidence I have in the database being to handle things like an I/O issue. In particular, if you have some sort of issue that requires the assistance of Support, then you reduce the risk of getting into one of those "Its their fault! No, its their fault" arguments between vendors.
Others welcome to add their thoughts, but in my Oracle experience, I've had good experiences with environments where ASM was doing all the redundancy, and just as may good experience with non-ASM doing the redundancy.
If you have to play with the new gear, maybe configure both options, run some benchmarks and make sure you don't hit any weird edge cases particular to your environment.