Can I change raid type from "raid4" to "raid-dp" on FAS3020 runing Ontap 7.3.3. without data disruption using the cli ?
Assuming you have a spare disk to become the "DP" drive then yes this can be added and removed on the fly.
If you dont have the disk then unfortunately you have to migrate off destroy the aggr then recreate.
following up on this thread, I have a FAS2020, which has a spare disk already, and is in RAID-4. When I try to change to "dp" it warns me:
Also, my aggregate has 5 disks + 1 spare. Shouldn't I be able to change to "dp" ?
You should as long as you still have enough spares left after this. Converting to RAID_DP consumes one spare per raid group, and you most probably have just one anyway.
Remember you'll still need a hot spare available after you've converted the aggregate to RAID-DP. If you only have 1 spare today, you won't be able to do this without buying more disks I'm afraid. Well, you can, but you'll have no hot spares and I'd rather have RAID-4 than no hot spares. A disk rebuild is quicker and more efficient in most scenarios than doing parity calculations on each read. It's a fine line, but I'd stick with RAID-4 unless you can buy more disks.
RAID-DP over RAID-4 has no read or write penalty, arguably RAID-4 only has one parity calculation, but the write cache removes that as a real concern.
Most disk failures are proactively failed, and a disk-to-disk copy has little or no impact on the CPU. So my RAID-4 system will have very little performance hit on a disk failure. I can still suffer 2 failures, although not at the same time.
In a RAID-DP system I have 2 parity disks but no hot spare, so any failure, proactive or otherwise, will cause a parity calculation, and this will have to be done for EVERY read until the failed disk is replaced. When the failed disk is replaced I then have to do an entire parity recalculation as I have no disk to copy valid data from.
Keep in mind that my thinking above is only valid for a small 2000 series with 12 disks where the RAID group sizes are also small. I'd vote RAID-DP every single time on a larger system without fail.
Because I have seen reconstruction failures caused by latent sector failures, and my priority is data availability above data access speed ☺ Such failures are probably less likely on NetApp due to the way data is placed on disks … but old habits die hard ☺
Proactive disk sparing requires 2 spares so it is out of question here anyway.
When I say proactive disk sparing, I mean when a disk is failed due to the quantity of bit level or software errors, rather than catastrophic disk failure. This doesn't require 2 disks, as far as I remember, 2 disks are only required for the maintenance garage which will run low level checks and reformat disks and bring them back into production based on software errors. But I could be completely wrong there