September 24, 20241 yr Hey people! I'm the owner of an AsRock H670M-ITX/ax which has 4 SATA ports, 2 NVME slots and one x16 PCIE slot (link to the Asrock site if you want to know more: https://www.asrock.com/mb/Intel/H670M-ITXax/index.asp). Actually my server has: - 2x12TB IronWolf NAS disks (1 parity + 1 storage) - 2x2TB SSD disks configured as btrfs mirror pool - 2x500GB NVME disks configured as mirrored cache so, as you can see, all my SATA slots are used but I need to add more disks. I actually have one disk I need to add to the array so I'm searching for suggestions on how I should proceed. I'm thinking about an NVME adapter with 6 SATA ports on it, like this one: https://www.amazon.it/Mengtech-Adattatore-SATA-NVME-espansione/dp/B0CZ3QR5HH/ but there will be a "problem": I will need to destroy the cache mirror and use only 1 NVME disk as cache device, which will mean no replication (aka data loss when the NVME drive will fails), and I don't even see any how to in the online manual site regarding this operation. In addition I think I should mention that each of the 6 SATA ports on the adapter, if they will be active at the same time under certain conditions, it could generate a bottleneck (I know it's unlikely to happen, at least in my env, but I should consider it anyway). so my questions are: is this approach doable? using these type of adapters, will it still be possible to spun down the drives? Is there any better alternative to this solution, keeping in mind that I would like to be able to expand my array even more in future? Thanks for any feedback
September 24, 20241 yr Community Expert I am testing with what looks like that exact device (here) on an ASrock 780 It seems to add 1-2W of energy consumption if installed but not hosting drives. So that's nice. Using 4 Exos HDDs, I am not noticing any bottleneck so far. I do recommend you verify the firmware on it as they have been known to ship very out of date. Here is the tool and an with an older update link. Attached is the most current ASM1166 firmware I could find, Nov2022. Also, when plugging in SATA cables, either do that before installing the card on MB, or be sure you are supporting it when plugging it in. The backplane wants to bend a lot when pushing on it. 11180000.rom
September 25, 20241 yr Author Thanks! I bought it, so next week I should be able to test it. So now the next question is: shoud I rebalance my nvme cache pool to "single mode", using the btrfs function to be able to remove one of my 2 nvme drives? Is that correct?
September 25, 20241 yr 34 minutes ago, quietwalker said: So now the next question is: shoud I rebalance my nvme cache pool to "single mode", using the btrfs function to be able to remove one of my 2 nvme drives? Is that correct? No, single mode aggregates the space of the drives, interleaving writes where possible for speed. You need RAID1, where the drives are duplicated, so you can then remove 1 and still have your data.
September 25, 20241 yr Author thanks Jonathan! I guess the other alternative could looks like: - set all shares to put data in the Array - stop docker and VM services - run the mover - check from logs that everything moved into the array (may be worth a doublecheck from terminal too) - destroy the cache pool - remove 1 nvme disk - recreate the cache pool with only 1 disk - rollback share setting to put data in cache again (for those who need cached data) - execute mover again (should move back in cache all the necessary stuff, like appdata) - start docker and VM services again does it sounds correct too? Anyway I think I will run unraid in single disk cache for a month or two and then I will create a new cache using SSDs instead of nvme. In that way I could potentially buy another m2 to sata adapter and mount something like 14 sata disks. EDIT: I added above the missing steps @JonathanM mentioned in the next comment. Thanks! Edited September 25, 20241 yr by quietwalker
September 25, 20241 yr Solution Only thing I see is before you run the mover, all services that access files on the pool need to be shut down. Typically that means disabling both the docker and VM services, not just stopping the containers or guest OS's.
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.