Confirming Setup Solutions & Approach

[FerrousEULA]

Hi Everyone!

Context:
I'm new to TrueNAS and ZFS. I'm migrating from a Synology DS413 because it's slow, the WD Blacks inside of it are at 60k hours, and it's time to make changes to ensure the safety and flexibility of my data. I am also dabbling in a self-hosted Plex, Home Assistant in a VM, etc.

I've read the primer, run my own experiments with spare drives, and am currently backing up all of my data to the cloud to prepare for a final migration before then selling off the DS413 device itself. I am trying to keep costs to a minimum, but will come back to this project periodically to continue improvements.

The Ask:
I'm hoping some of you can confirm my understanding and critique my planned solution before I go too far down this path.

The Current Setup:
4x2TB WD Black (SHR 1 drive fault tolerance, effectively RAID5) in a DS413. They're at 60k hours and going strong, but it's time to do better.

The Destination:
Desktop Tower, Core i5 2500k, Z68 Extreme3 Gen3, 16GB DDR3. This computer is roughly 10 years old, runs fine, tests fine. I've already set this up on Scale and run experiments. All behaves stable.

Additional Storage Devices:
2TB WD Black, 1TB HDD, 128GB SSD (Boots Scale), 256GB SSD (Holds Apps), 2x12TB Manf. Recert. Western Digital Ultrastar DC HC520 on the way.

The Target Solution:
3x2TB RAIDZ1, 2x2TB Stripe (Plex) | Both backed up to 2x12TB Mirror w/ scheduled snapshots as well.

I intend to set up a 20% size reserve dataset on each pool in order to protect the operational headroom if datasets fill up.

The Transfer:
Backing up all data to cloud now, then copying to the 2x12 mirror, then pulling down the DS413 and moving drives over, then reformatting those same drives as the front end of my solution (that sees more active use) until they die with the Backup mirror as a safety net. I would keep 1TB of the most critical data in cloud storage just to be extra safe.

My logic is that if the 3x2 RAIDZ1 dies on me I'll just take a drive from the 2x2 Stripe to resilver, and if it dies while doing that I still have the mirror backup.

What am I missing? Does this approach read like I absorbed all of the necessary information to do this responsibly?

Addendum:
If so, I would have some follow up questions that I will continue searching in the meantime.

1. The 20% free space target is per data set, not per pool correct? If so, then the 20% reserved space just exists to expand dataset quotas temporarily to allow cleanup operations?
2. Is there a reason not to set up scheduled snapshots to a drive other than the boot device? I didn't see how to do so immediately, but am confident I can figure it out, but a quick explanation would be welcome if you're feeling generous.

Thanks for taking the time to read and support this eye opening journey of proper data management!

 

[farout]

Hi

I understand that you want to re-use your old hardware and save costs. However you also wrote, that you want "to ensure the safety and flexibility of your data", running VMs, Plex etc.

While your i5 supports Vt-x, it does not support Vt-d, is now 13 years old (not 10) and has a TDP of 95 Watts - basically a space heater. It also has only 4 threads (no hyperthreading). No support for ECC. The old gaming motherboard might also give out in the near future.
Also you are choosing ZFS, and then take a major advantage away from it by starving it with only 16GB of RAM.

Will it run SCALE ? Yes, but dont expect "safety and flexibility" from it.

 

[FerrousEULA]

Hi

I understand that you want to re-use your old hardware and save costs. However you also wrote, that you want "to ensure the safety and flexibility of your data", running VMs, Plex etc.

While your i5 supports Vt-x, it does not support Vt-d, is now 13 years old (not 10) and has a TDP of 95 Watts - basically a space heater. It also has only 4 threads (no hyperthreading). No support for ECC. The old gaming motherboard might also give out in the near future.
Also you are choosing ZFS, and then take a major advantage away from it by starving it with only 16GB of RAM.

Will it run SCALE ? Yes, but dont expect "safety and flexibility" from it.

Thanks for the info! Fortunately, last night I decided against the VM - Home Assistant isn't necessary for my purposes yet - so I wouldn't be running it all the time. It's worth noting that should the machine die I can wait a few weeks while I source a new mobo/cpu/ram, etc. Any recommendations you'd have there would be great.

I suppose the question I'm asking is: what is the real risk of losing everything, as opposed to just losing one of the two copies present on the machine at any given time. My understanding was that data being written at that time could be damaged, but that it wouldn't destroy the entire system (different pools) as well. Is that wrong?

 

[farout]

I am not an expert, that could tell you to what extend ZFS can selfheal corrupt data. It sure will tell you that it found errors and that you might have to recover from backup.

If your mobo is dying, it might corrupt data beeing sent to the backup drives. Then you would have to rely on your primary datapool- which consists of 60k hour drives.

I assume your experience with your current NAS was pretty painless and trouble free. And I assume you want something similar now.

Thats why your "new" setup, that consists of old consumer hardware, and backup drives of unknown heritage raises some red flags, that can result in downtime and hassle as you will have to juggle data to a - for the moment - healthy pool.

 

[FerrousEULA]

Appreciate your feedback. I'll do some digging around about a cheap mobo/cpu adjustment that could be better.

Ultimately I liked Synology for its stability, but I don't like the idea of being locked into its proprietary raid format or its really slow performance. I'd have to wipe the data to reformat and I don't feel its worth the hassle. The performance from this TrueNAS setup is already a huge improvement - it sounds like I just need to do some more planning around stability risks with those old equipment.

I've been thinking I'd sell the DS413 and then use those funds to make other improvements to the TrueNAS box. Ultimately all of my most critical/valuable data will still be backed up to the cloud.

That would give me 1 drive fault tolerance pool, backed up by a mirror pool, all backed up to the cloud. In theory that should be enough to protect my data while I iron out any performance kinks and settle into my new custom NAS setup.