Virtual Disk with Real disks in RAIDZ1 for Truenas

[welcometors]

I have one primary disk for running VMs (3.8TB Samsung pm9a3 u.2) and two 1tb nvme SSDs (Samsung 980 pro). I want to create a Truenas core/scale VM using 1TB virtual disk from primary disk and 2x 1TB nvme in Raidz1 to get 2TB usable space from my NVME drives. My usecase is just to use them as a data drives with redundancy for NextCloud.

Is this a good idea? What about the write amplification from the VM using virtual disk in raidz1?

 

[leesteken]

using 1TB virtual disk from primary disk and 2x 1TB nvme in Raidz1 to get 2TB usable space

66% usable space is more difficult than it looks: https://forum.proxmox.com/search/6587155/?q=raidz1+padding&o=date

 

[welcometors]

66% usable space is more difficult than it looks: https://forum.proxmox.com/search/6587155/?q=raidz1+padding&o=date

So is it better to go with LVM pool using both NVME SSDs with Casaos or something lightweight for network sharing and instead of going for redundancy go with daily backup? I want my NextClould storage to be as reliable as I can

 

[leesteken]

I want my NextClould storage to be as reliable as I can

For reliability use a three-way mirror, so that your data is still redundant when you replace a broken drive.

So is it better to go with LVM pool using both NVME SSDs with Casaos or something lightweight for network sharing and instead of going for redundancy go with daily backup?

I like ZFS because of the self-healing and bitrot detection but raidz1 is not the same as RAID5 and ashift, volblocksize and padding need to be balanced to get the most space (and cannot be expanded later). Backups on a remote system and on different media is always a good idea.

 

[welcometors]

For reliability use a three-way mirror, so that your data is still redundant when you replace a broken drive.

I like ZFS because of the self-healing and bitrot detection but raidz1 is not the same as RAID5 and ashift, volblocksize and padding need to be balanced to get the most space (and cannot be expanded later). Backups on a remote system and on different media is always a good idea.

Thanks for the understanding my case. Currently I don't need extra space and I can go for higher capacity NVMEs in future.

So the best way to run NextCloud is to:
1. Use primary disk for the VM install (debian with Casaos)
2. Use 3 way mirrored zpool from 3 NVME as data storage and pass it to VM
3. Install NextCloud as docker and use second disk only for storing data
4. Daily back-up data disk to my Synology NAS + pcloud?

 

[leesteken]

Is it useful to run NextCloud on NVMe? Is it not mostly limited by (external) network speeds? If you do regular backups, maybe you don't need drive redundancy? It's not clear to me what you mean by "best way" or "most reliable". Maybe other people here do have experience with NextCloud and advise you better.

 

[welcometors]

Is it useful to run NextCloud on NVMe? Is it not mostly limited by (external) network speeds? If you do regular backups, maybe you don't need drive redundancy? It's not clear to me what you mean by "best way" or "most reliable". Maybe other people here do have experience with NextCloud and advise you better.

Yes, I'll be mostly limited by my internet speed (200mbps). However my server (Minisforum MS-01) doesn't have space to put hard disks. And it doesn't have ECC memory. So at least protection against bit rot will be good.