SOLVED First time build and a comprehensive plan

[Maxburn]

OK, I've done a lot of reading and I think I know what I'm doing but one last minute reality check.

Use case, this is a general file share dump. Plenty of movies played back with a Dune network player, no transcoding. The most stress this will see is when something gets written to it and scheduled rsync tasks. Somehow over the past couple years that task has been enough to kill a DS1515+ and DS418 through Synology bad luck. I want a set it and only do occasional easy updates machine.

Supermicro X11SSM-F-O and manufacturer link
Intel Celeron G3930, has ECC
Single 16GB DDR4 ECC, from supermicro approved list.
Seasonic S12II 620W PSU
SAS 9207-8I based on recommendations seen here.
Boot from USB via Sandisk ultra fit 64gb
Decade old Yeong Yang cube server case with a couple new bay adapters. Should hold up to 18 3.5" drives.

Main pool (in above machine)
Existing four 8TB Seagate Ironwolf will be the main pool that everyone reads/writes to. Drives are approaching a year old. I'm thinking about making this 4x1 ZFS1 but am toying with 2x2 ZFS1. Calculator says that's either 14TB or 20TB useable but I'm not sure which way I should go with it, seeing mixed advice here. I'm at about 8.5TB in use but slowly growing. There are a couple automation tasks downloading files to this at any time and three computer using it as a backup, this will be on 24x7.

These drives are currently in a DS418 with a crashed volume after a DSM software update. Disk 1 and 2 have logs saying write failures but SMART says they are all fine. If I have to RMA two drives so be it, but I've mentally written off my data here.

Complete Backup Pool (in above machine)
I was thinking about skipping this but realized I have eight existing 2TB Seagate LP(PDF) laying around so why not use them? This is the reason I added the LSI card. These drives are 7+ years old, some are already RMA replacements but they will all go through spinrite before going in the pool. I would like this pool to sleep when not being used (noise and power concerns) and have read that to accomplish that these need to be the ones attached to the motherboard though ideally I'd prefer my main pool attached to the southbridge ports. I figure I'll set a rsync task from the main pool to this one every week or so, otherwise they should be off. Should be 11.8TB useable space in a 8x1 ZFS1 pool.

Critical cold backup (attached to above machine)
USB3 Seagate 8TB external. This will hold the irreplaceable data from above main pool, but on a longer backup schedule maybe doing a rsync task every month or so just in case of malware or cryptolocker etc. I will have to remember to attach it and run a rsync task, hope I can do that. Will be looking for another one on sale over black friday, this one is full and stopped doing weekly backups last month which puts me in a double bad position with the crashed DS418. Currently holding a second best bet of my data, but a little old.

Off site backup
Synology DS1515+ with SHR1 across three 2TB Seagate LP (also ~7 years old) and two 8TB shucked WD Easystores (6 months old) that yielded Red drives getting me about 12TB SHR1 formatted space. The DS itself is back from RMA about 6 months ago. This is (and was) a nightly backup target from above main pool over VPN. This currently has the best copy of my data, so looks like the above machine will go for a drive when it's built as downloading 9TB over a cable modem upload speed isn't doable.

 

[Jessep]

Hardware looks fine other than a minor concern about "bay adapters". Sometimes they aren't the best quality, usually don't support hot swap, and can introduce strange/random errors. Your link to them isn't working for me so I can't offer an opinion about that model.

Standard thought is any vdev using drive sizes over 2TB should be in Z2 or Z3 due to the longer resilver times and the risk of a second drive dropping during resilver. Keep in mind it's not a straight 2 drives "parity", it's also keeping your pool under 80% full (to prevent performance loss). It would make sense to add 2 or more additional drives from the start to cover data growth.

Backup pool using older drives also sounds fine, this will use a fair bit of power. Drive tech has come a long way, at the least do the calculations on power usage/cost in your area, you might find that you can pay off a newer larger set of drives in a reasonable timeframe.

SMART short, SMART long, Badblocks, SMART long seems to be the suggested testing before into production with scheduled SMART short, SMART long, and scrubs in production.

 

[Maxburn]

Hardware looks fine other than a minor concern about "bay adapters". Sometimes they aren't the best quality, usually don't support hot swap, and can introduce strange/random errors. Your link to them isn't working for me so I can't offer an opinion about that model.

I agree. This is a link to what I actually purchased with some decent pictures. My main concerns with things with back planes are they introduce a new point of failure for signal integrity, but these don't have a back plane. It's literally just a way to hold a lot of drives similar to internal bays. No hot swap stuff to get in the way.

Standard thought is any vdev using drive sizes over 2TB should be in Z2 or Z3 due to the longer resilver times and the risk of a second drive dropping during resilver. Keep in mind it's not a straight 2 drives "parity", it's also keeping your pool under 80% full (to prevent performance loss). It would make sense to add 2 or more additional drives from the start to cover data growth.

I'll think about it, based on my backup sizes I seem to have room to spare on the main pool. Thanks, I hadn't considered recover times in the event of failures.

Backup pool using older drives also sounds fine, this will use a fair bit of power. Drive tech has come a long way, at the least do the calculations on power usage/cost in your area, you might find that you can pay off a newer larger set of drives in a reasonable timeframe.

I actually did the math and even if the backup pool is on 24/7/365 it will cost me about $45/year with my rates. That isn't too bad but I'm more interested in not having the noise which is why I'm wanting to spin them down while not in use. Does this actually work with FreeNAS?

SMART short, SMART long, Badblocks, SMART long seems to be the suggested testing before into production with scheduled SMART short, SMART long, and scrubs in production.

Good advice, thank you.