To help determine is it worth it:
1. What are my availability requirements? RPO, RTO?
2. What resources you are consuming to get that extra availability
3. How much more availability are you getting for it and is that worth it
4. What it would take operationally to recover if it's not automatic..
Things along these lines should get you to your answer.
If you are SAN booting the only thing that presenting 2 LUNs to the host and mirror would prevent is in an event that one LUN goes offline for some reason. The availability that comes with mirroring disks that are onboard the host are possibly three-fold: SATA/SAS path failure if there is only one path available to the disks (not usual), ASIC failure (if you have more than one RAID card), and disk failure. Availability for the SAN boot LUN comes from RAIDDP, redundant HBA paths to that LUN, redundant controllers, and redundant paths to all drives in the system.
If you are extremely paranoid you could setup a mirrored aggregate for the boot drives, snapmirror to another local array, or a metrocluster.
In my estimation it would not be worth it, but again depending on your availability requirements for the boot drive it may be the right decision.
@netappmagic wrote:
With NetApp volume/lun provisioning to Unix hosts, are there needs to mirror root and bootable lun on hosts side?
Snapshot can make up data loss, and the storage can provide ways to protect data, but in terms of protecting bootable lun on the host, I think we still should create the mirror to root.
What would you think?
