Intel
Explorer
- Joined
- Sep 30, 2014
- Messages
- 51
I'm seeking advice on the best approach / pool setup for my new NAS. I have always had a striped vdev mirrors configuration of 4 disks on my current/old NAS.
My new build has larger drives and NVME and I wanted to get as many IOPS and speed as possible while having some backup assurances (I haven't had a hard drive die on me in the past 3 years). I am considering having a "speed" zero redundancy pool and then a backup "slow" pool.
my hardware:
If the above is true, and stripe of mirrors don't hold metadata in spinning drives (aka there is no backup to whats on the nvme in a mirror stripe). I can see another way of achieving my goals, if I am willing to give up some data redundancy/parity.
speed_pool
- single data vdev (stripe of 2x18TB disks)
- metadata + SLOG + ZIL in nvme disk
total ~33TB raw storage + should be super fast IO. Risks: if any disk or nvme dies everything dies.
slow_pool
- single data vdev (stripe of 2x18TB disks)
- TrueNAS hourly scheduled backup job that syncs changes from speed_pool into slow_pool
Are there other alternatives I should consider, or any other recommendations? Is my understanding of metadata storage on spinning rust drives on stripe vdevs with nvme ssd incorrect, would nvme dying kill the entire pool if it were a 4x18TB + nvme setup?
My new build has larger drives and NVME and I wanted to get as many IOPS and speed as possible while having some backup assurances (I haven't had a hard drive die on me in the past 3 years). I am considering having a "speed" zero redundancy pool and then a backup "slow" pool.
my hardware:
- 1TB PCIe 4.0 nvme (3500mb/s rw - passthru to TrueNAS VM)
- 4x 18TB drives
- TrueNAS Scale (running as a proxmox VM)
- SAS3008 HBA (passthrough from proxmox to VM)
- Proxmox VMs
- Filestorage
If the above is true, and stripe of mirrors don't hold metadata in spinning drives (aka there is no backup to whats on the nvme in a mirror stripe). I can see another way of achieving my goals, if I am willing to give up some data redundancy/parity.
speed_pool
- single data vdev (stripe of 2x18TB disks)
- metadata + SLOG + ZIL in nvme disk
total ~33TB raw storage + should be super fast IO. Risks: if any disk or nvme dies everything dies.
slow_pool
- single data vdev (stripe of 2x18TB disks)
- TrueNAS hourly scheduled backup job that syncs changes from speed_pool into slow_pool
Are there other alternatives I should consider, or any other recommendations? Is my understanding of metadata storage on spinning rust drives on stripe vdevs with nvme ssd incorrect, would nvme dying kill the entire pool if it were a 4x18TB + nvme setup?