I will likely have everyone beat.. mostly because nobody would keep this ‘trash’ around
I am running 25.10 on a .. wait for it..
Core 2 Quad 6600 quad-core Xeon CPUs, released on November 2, 2006 with JUST 8GB of RAM
I use it as a replication target and it saturates gigabit ethernet with a pool of 3 ancient 2006 seagate 1.5TB drives in a raidz
yep folks.. nothing fancy..
Dangit. You’re gonna make me pull out my G5 PowerMacs or Core Duo iMacs and Mac Minis and figure out some way to run ZFS on them…
Yes! That is exactly what I was thinking of when I started this thread! I keep telling people not to care about specs too much. If you have it and it runs; It is fine.
I have a G4 original Mac mini I pull out and boot one in a while running snow leopard server edition
ZFS is NOT that resource intensive (truenas middleware bloat not included) but can be if all the bells and whistles are turned on
ZFS is improperly used by almost everyone that does not have storage admin training.. and those that have older, hardware raid traditional storage training are worse - software defined storage is a different cat
ZFS in NOT A BACKUP REPLACEMENT, RAIDZ and mirrors are to permit a storage box to remain online during a failure, while a backup is brought up and/or remediation is made.. not a replacement for said backup, not a protection for data loss, but from bitrot, 2 different things entirely - 3 2 1 backups still required although after nearly 2 decades of running ZFS have not lost a file yet.. knock on wood
ZFS only needs gobs of RAM for DE-DUP which the vast majority of people who run it dont need it.. dont understand its original purpose - and yes.. while more ram permits more active data to be cached, for many running a file system many times the data is never requested twice so caching has again only advantages for workflows continuously accessing the same data
LOG and special vdev devices are overkill in almost all situations, add cost, complexity and points of failure, some catastrophic, and if you need a ‘fast’ pool, just about any SSD or NVME pool will suffice
when hardened properly, ZFS is quite possibly the most cyber secure storage solution and due to its sun/oracle roots and longevity one of the most stable
ZFS being software defined has many dial, buttons, and switches for configurability and customization to workload, it doesn’t mean you need to turn them all on and use them all - the right options for the right job..
Did you mean to post that here? I don’t get the connection.
Edit: Maybe I get the connection. Never mind. Anyway, I agree regarding resources. Hence the thread.
My oldest system is still running TrueNAS CORE.
It’s a Supermicro SuperServer SYS5018D-MF paired with an Intel i3-4160.
It still runs great and still in use today though not for 24/7 use. It only gets turned on as a replication target.
yes.. the connection is simple..
Confucius would say.. dont kill a mosquito with a canon ..
what I am saying is that MOST zfs hobby users ie truenas community members .. build canons on way over-spec hardware, add un-needed zfs features and complexity when not needed or fully understood thus wasting resources or claiming that ZFS is ‘heavy’ and requires lots of cpu, ram, exotic ssd/nvme etc.. when really they just need to learn how to build out a zfs instance appropriate for their workload
if that makes more sense
I personally just keep it simple. Striped mirrors with zstd compression and that’s all.
The only thing I made sure I “overspec-ed” are ECC RAM and enterprise SATA SSD’s for block storage used to host VM’s.
Personally, I don’t understand why people bother with encryption at all. 90% of home users don’t have data that state actors would be interested in that warrants it. And the few things you do want to encrypt (like password stores) are often already stored encrypted by the password manager anyway or you can just explicitly encrypt just for that use case.
More often than not on the forums, I actually see encryption become their own downfall where they either forget the key or misplace it, etc. and their NAS effectively turns into self-imposed ransomware.
If i may say, i would not consider ECC an over spec… IMHO instead something where Is worth put budget (but i repeat, IMHO).
More than see people bother with encription, i have see more people
There’s usually a tendency to overspec hardware when buying it (not only related to Nas), thinking that a bit of extra performance will help it last longer over the years.
But IMHO you’re just spending more today on something you don’t actually need, and that you’ll probably be able to get for much less later (btw, if you really need It).
An exception could be made when the hardware is very inexpensive… because as we’re experiencing now there can be significant price fluctuations that might prevent (or delay) an upgrade in the future.
Hmmmm…
Sometimes, best laid plans are wrecked by reality. I purchased a more expensive embedded motherboard way back when with the idea that I’d be able to run ZoneMinder and VMs on CORE in addition to the NAS. Had I bought the less expensive motherboard, I would have saved $500 and it would have allowed me to run even faster SMB transfers since the -2C- version of my motherboard runs 500MHz faster than the one I have.
But while cheap NVRs can record camera images all day long on ancient CPUs, the combination of ZoneMinder, CORE, and the D1537 resulted in dropped frames, unreliable operation, etc. My experience with Frigate was not better, that package, while cool & the new hot thing, was also not reliably recording. More missed events, and a parser that would have delighted my former sadistic music teacher with its unending pickiness re: syntax.
I don’t disagree that folk frequently over-provision but we should also not lose sight of plans hitting reality. Some plans survive that encounter, some don’t. With the benefit of hindsight, I would now prefer a bare motherboard with the ability to fit exactly what I need vs. the embedded route I chose.
That’s not a criticism of the motherboard (it’s been working fine) but rather an acknowledgment that all-in-one solutions come with drawbacks, such as not being able to fit a more modern, more-power efficient CPU in the place of the embedded Denverton era unit I have now.
I’d like to thank everyone who participated in this thread with their systems already. It showed that the recommended specs are just recommendations and that TrueNAS can be run on much older and weaker hardware if the owner is willing to sacrifice some performance or extra features. Or just use it as a NAS and not an All-In-One-Server. I should have called it oldest/weakest hardware.
Looking forward for others to join the inverse e-peen contest 
i’m running a Proliant Microserver Gen 7, Originally bought as a stock N40L, outfitted with 5 2TB HHDs, replacing the CD Drive with one more HHD than the system was designed for, one 32GB Thumb drive for the OS, and 16GB of ECC RAM. All bought with the best suggestions of the FreeNAS community team suggestion. As a Linux user, I was sort of “experimenting” with BSD as the few BSD hobbyist fanatics I was acquainted with were always telling me “BDS is true UNIX and Linux is not.” I couldn’t argue with them as, in the office, we were all using HPUX and SUN Microsystems (as well as old IBM 370’s, going further back. I remember when a 300 baud modem waq considered “state of the art.” Anyways, because I knew I had to keep replacements around, I started with FreeNAS 7.1 and ZFS Z1 single vdev. My HHDs generally ran for about 5 years before I’d get the yellow triangle on the dashboard which required downing the machine for a couple of hours to replace the failed drive (as far as ZFS reported it), and things would get going again. The system was set up to store my ever growing collection of Korean dramas my wife wanted to watch to put her to sleep every night, so the down time wasn’t really noticed by her. I’ve scoured ebay for spare and adjunct parts, my forst purchase to be a 5 bay HHD box with a port multiplier. I created a second vdev on it and installed another 5 2TB HHDs as I was nearing “full” (80% of capacity). What I didn’t like about this setup was the eSATA connection would get bumped, causing a failure, and scrubs took about 27 hours. Around 2016 I began replacing my 2TB drives with 4TB drives as the older drives failed (FYI, I tended to always keep at least 2 new spares in boxes on my home office floor, buying them when they generally reached a price of around $85 for the most TB I could drive. Eventually, around 2021 I had only 2 2TB drives left that hadn’t failed but had at least 5 years of 24/7 running time on them. I replaced them and took the big step of resizing all 10 drives from 2TB to 4TB, doubling my capacity. Then I bought 2 14TB drives on sale and took the risk of creating two Z1 replications of my entire ZFS file sysstem (Oh yes, I had also replaced the Thumb drive with an eMMC drive with a SATA to USB adapter, backed up by a small external SSD with USB adapter for the OS.
Now the time to rebuild my filesystem as a Z2. This went smoothly, but I began to worry about parts of the MicroServer failing. Back to ebay, where I found a refurbished N54 MB $100, and now, 350 watt power supplies designated built for the Gen 7 MicroServers.
Still beleaguered by the 27 hour scrubs, I began to took into replacing the 5 bay eSATA external drive box with something run through an HBA. Looking around Newegg, I saw a 4 bay SAS/SAT box for $79 and envisioned stacking two boxes together, running them through the 350 watt “spare” power supply I had in myb “spares.” I took the risk and bought them as well as a Chinese HBA LSI9000 8e card as I had to stay in the realm of pcie2) through ebay again, as well as appropriate external SAS cables. I won’t get into the learning curve of SAS terminology here. Eventually it all worked, and the Chinese cables suffered what we must all know - poor QC, so I had to buy replacements as RMAs back to China don’t always work like we could like them to.
Finally, I began to think how to modernize more, deciding to upgrade the original SATA2 controller built into the southbridge of the Gen7 MBs, realized I could run the HBA into a SAS multipliier, take the HHDs off the Microserver backplane and put them onto the SAS multiplier. This took finding a SAS multiplier with the appropriate fittings and doing so. Result, my scrubs now take 3.25 hours.
Oh, yes, I must mention, when I got the N54 MB, I pulled the heat sinks off the CPU and SB, cleaned the crusty Thermal Paste, before installing it in the Proliant box (so I now have an N54 instead of an N40) and I also cleaned the still working N40 MB heat sinks that is now being maintained as a working spare. One thing I might mention, as there are still more Gen 7 users out there than I tought, is I had to go through a lot of “dead ends” trying to find the appropriate patches to the BIOS and pictures of disassembling the Microserver box.
Send me a message if you want more explicit information on my journey. I want to thank the FreeNAS and TrueNAS pioneers who have taken me through this journey with explanations of WHY things are done the way they were and are. Even though I’m ore familiar with Linux, I will continue to use CORE for my media server until I see comments from the SCALE users that suggest that all the problems the Linux community has had with adopting ZFS into Linux. I will say that once a year I replicate my whole media collection onto an external HHD (via the eSATA port, but next time through a spare slot inside my Microserver since it will run at SATA3 speeds rather than the old SATA2 via the southbridge chip) which gives me the ability to keep my wife’s hobby of watching her dramas through the night via my desktop running ARCH Linus with ZFS drivers provided via dkms.
Wow - some old iron out there.
My servers are HPE DL380G9s. They are rock solid and I have 3 of them - absolute overkill.
Some day I will migrate to SCALE. But for now, CORE works just fine (1 server is just TrueNAS, other 2 are running ESXi 7).
Great to see all the old gear still running like champions!!
The other day I looked at the Zabbix Dashboard I configured and found out that my Dell T330 had a bad memory module. The error has been existent for a few days until I have seen it, but TrueNAS SCALE didn’t bother. It just had less RAM than before. No services have been compromised… ECC is worth it…
I have several desktop cases that serve as small task machines. TrueNAS Scale was running on my 1st Gen i3, 8gb ddr3, plain intel board, 2tb WD purple, Patriot SSD (256gb 2.5”) without dedicated video, ran it headless and running Linux Mint as an OS for files. Lost my DVR (20yr old unit with 960h analog cameras) so looked into a self created alternative platform. Moved to TrueNAS Scale on the advice of a close friend, then added Nginx and Frigate. Grabbed a POE switch and a few older IP cameras and was off to the races.
And just like that, one of the 10TB drives on my “lab” box threw a couple hundred errors and TrueNAS/ZFS failed it. Replacement ordered and should be here in two days. All critical data already cloned (replicated) to the production server and now I’m migrating the VMDK’s of the few remaining non-essential (but don’t want to have to rebuild them) VMs to the second pool of disks in the machine.
(All goes to the point of one of the earlier posts about 3-2-1 and ZFS is a tool to help keep things working while you’re fixing the broken bits. I’ve got the 2-1 part down; still working on a third copy for everything important.)
When you say hundreds, is it the classic UDMA CRC Checksum error that can totally spam? I ask because that error is usually related to cabling / connectors, not the drive.
It was at 8-9 errors the other day. This morning it jumped to 170, then 289. I’ll have to check it. You’re right, it may just need to be powered off and checked to see if something came loose.
At the very least, if it fixes itself after a scrub, I’ll have cold spares in stock.
I now try to use deOxit on every connector before I make the connection.
