TrueNAS Scale on Arm (2024 Thread)

Just wanted to make sure Jeff’s geerling’s thread was not forgotten.

Truenas scale on arm would be a perfect.
pi 5 with the nvme bases\hats and even the pi4 should be more than capable to have scale in it.
been running a homelab k3s cluster for about 2+ years in some pi4, the moment truenast supports arm i will insta upgrade
and as mentioned even VPS providers started to offer arm, specially hetzner.

2 Likes

To add to the discussion: one major complaint has been the lack of acceptable hardware. With the advent of actually decent PCIe in even base-level SBCs like Raspberry Pi 5 and RK3588 solutions like the CM3588, SATA and NVMe storage support is much simpler and more reliable.

There are now multiple SBC options with 8, 16, or even 32 GB of RAM, with CPU speeds on par with low-end x86, which is more than adequate for 2.5 or 10 Gbps NAS performance.

With higher-end Arm CPUs like those from Ampere supporting ECC RAM, and being available to the masses now with motherboard bundles like this one from ASRock Rack—especially considering it supports vanilla UEFI via EDK II…

The arguments against support due to a lack of adequate hardware seem to hold less weight.

I am now running two Arm NASes with ZFS, my HL15 with the ASRock Rack board as my primary NAS with 120 TB of HDD + 30 TB of U.2 NVMe SSDs, and a Raspberry Pi 5 as my backup NAS, with 32 TB of SATA SSDs. The Pi is pulling snapshots from the primary with syncoid, and all the setup for this is available in the Debian packages.

It seems like all the elements are there on the software side, and IMO, there’s a huge opportunity to keep TrueNAS as one of the primary “if you’re gonna build a NAS, you’re gonna use this” options across architectures.

I’m perfectly happy with my homegrown NAS, the whole setup is only a few hundred lines of Ansible playbook… but I would’ve been happier to run TrueNAS Scale on it :wink:

2 Likes

Yes, ARM64 has improved quite a bit in the last few years.

However, I would say that a SBC, (Single Board Computer), with limited, odd or no expansion is not really suitable for TrueNAS. This applies to all Raspberry Pis, regardless of memory size, amount of storage ports or network port speed.

TrueNAS is intended to be an Enterprise Data Center quality software product based on reliable, Enterprise ready hardware.

That said, ZFS with Linux likely would work quite well on ARM64 computers.

And using an example of the only “real PC like” ARM64 computer, does not mean that enough system boards would make it worth while for software development.

iX does not have any incentive to purchase ARM64 SBCs, pay for software development to port TrueNAS SCALE to such devices, and then thoroughly test such for reliability. That does not even mention the expected extra features that SCALE has, VMs and Apps. All of which would not only be expected to be supported, but work as well or better on ARM64 as it does on x86_64.

But, their is nothing stopping someone else from porting TrueNAS SCALE to ARM64 on their own.

I am not trying to rain on anyone’s parade, yet the reality is that iXsystems is too small a company to make such changes. Yes, it would be nice if iX made several low cost home or small office NASes, that might be based on ARM64. (To reduce the cost or power consumption compared to x86_64…) But, the market at that price level is too competitive and fractured.

3 Likes

Some of it may work, but we (iX) have not tested.
If there is a serious effort to port and test, we’ll certainly look at any bug reports.
Join the discord channel.

2 Likes

TrueNAS on ARM would be nice, but is a bit of a strech.

What would be very nice, feasible and testable, however is an ARM version of TrueCommand, to manage our NAS from a Raspberry Pi or similar—no need for serious server-grade plateform here.

2 Likes

Or if someone happens to be running stuff on ARM, they can throw in TrueCommand. It should be more straightforward, as a containerized thing rather than a full distro.

Agreed.

Now this would be much more doable. The assumption is that TrueCommand is just a normal App, in the sense it does not have specific hardware requirements. I mean, yes, TC needs memory & CPU, plus both code disk space, and running disk space. Not to mention network. However, compared to TrueNAS X, (aka Core or SCALE), TC has very minimal hard requirements.

On the other hand, iX still has the limitation of what their Enterprise customers would use to run TrueCommand. So back to the question would it be financially viable to develop TC for ARM, when it is probably that few, if any, Enterprise customers would use the ARM based version?

Given SCALE’s numerous “issues”, I’d rather see those worked on first. Not to mention Core’s up and coming release…

Yes, TC is a normal container App. Anyone is welcome to try it on ARM, but we have no plans to focus on that. It may work for all I know.

Yes, 13.3 is the next focus. BETA is in QA now.

SCALE issues are certainly coming down. Dragonfish has 15K users already (fastest adoption ever). Performance at high end seems to be about 40% better than 13.0. I’d be keen to understand your top 5 “issues” to see how we are tracking.

1 Like

Er… Wouldn’t that require that the container is built/compiled for ARM architecture rather than x86_64?

1 Like

Correct, we’d have to spin up a specific ARM build / container to make that available.

The whole idea of a ARM variant of SCALE interesting and becoming harder to ignore. It’s just not something we have a valid business reason to go after hard right now. I would say its a good community project though, if somebody wants to take the current scale build repo and adjust things to see if they can do an ARM build that would be super interesting. If it gets traction we’d be happy to review and see what the delta looks like to make it a better supported thing.

2 Likes

Agreed, we’re already planning a .1 sometime in May. Expecting to knock out the biggest issues we’ve seen so far. If you have specific ones you think should be on that list, then please let us know!

No, I have no specific, (or top 5), issues to report for SCALE.

The comment was based on the amount & type of forum posts, both old forum and new forum, on SCALE issues.

However, a lot of that could be many more people coming to SCALE for Apps & Linux, that either have unrealistic expectations of the software. Or using less than ideal hardware for their intended use.

There is a reason why I wrote that SCALE Beginners Intro Resource. It appears that many new TrueNAS SCALE users are expecting something different than what they get / got.

One thing that is going to continuously bite SCALE software development, is that Linux people ASSUME full driver support for their new, wiz bang, do hicky that does STUFF, which requires kernel driver support.

For the old TrueNAS Core, (and formerly FreeNAS), the answer was easy. If the kernel driver did not exist in current FreeBSD used in Core, then here are the choices:

  • Wait for next release of Core
  • Build it yourself
  • Use some other NAS software

And if the kernel driver was not expected for the next FreeBSD used in Core, then the bottom 2 were your only choices.

But non-Enterprise Linux users are EXPECTING current kernels. And will complain loudly if they can’t use their new, wiz bang, do hicky that does STUFF.

1 Like

I actually tried that last year lol.
I tried to run the scale-build in a pi4, i spent a whole day fighting against the requirements/dependencies ( scale-build/conf/build.manifest at master · truenas/scale-build · GitHub )
i ended up giving up since there where plenty of dependencies only for AMD64
it was good attempt tough, i might try again someday in the future

Microsoft is releasing another dev kit now with the snapdragon elite x. Apple has been making their own silicon for years. I would bet most servers in AWS are running ARM. Nvidia’s Jetsons are ARM based. Nvidia’s suspected 2025 processor is now rumored to be built by Intel using off-the-peg Arm cores which is a wild rumor but it’s now come from multiple valid sources. I can’t imagine seeing Microsoft supporting ARM, most linux distros having some level of support for ARM, and TrueNAS not seeing the writing in the sky.

Yes, that is understandable.

However, you missed the point of NAS hardware. Yes, lots of desktop hardware and laptops are being released with ARM64 CPUs. Their are even some small ARM64 servers that are retail available. But, real NAS hardware appears, (note my wording, appears), to be either quite limited or non-existent for ARM64.

Definition of “real NAS hardware” could be some of these things:

  • Ability to support >32GB of memory
  • Suitable NIC vendor, (aka not Realtek Ethernet chips)
  • Wired networking, (home or small business could get away with WiFi but wired will beat WiFi in reliability and speed every time)
  • Enough storage ports, (like >=6 SATA, and or >=2 NVMe)
  • Reliable power supply, (do you think a laptop style power supply should be considered server grade reliable?)

While I wish ARM64 and even ARM64 NASes well and a good future, I just don’t think this is the year we will see them. Maybe next summer.

Up to 2 TB of ECC RAM. Intel NIC. No SATA port but plenty of PCIe lanes for HBA(s) or tens of NVMe drives. Available. :star_struck:

I admit it is somewhat overkill for a home NAS…

1 Like

HPE does; that’s what they equip their current Microservers with.

If a decade ago EV companies said “electric vehicles won’t work because there are no chargers” then stopped building…we’d have no electric vehicles. The increase in EVs is what incentivized the creation of charger companies like ChargePoint and Blink. Apply that to this, if the software is there and the gap will get filled by companies who demand using ARM for everything and the board manufacturers will adjust.

I agree that TrueNAS shouldn’t be doing this because of lowly peons like me running Raspberry Pi 5 NAS( thanks @geerlingguy ) but TrueNAS has a chance to take that step so a hardware company can build for it. TrueNAS would then be the mature offering for ARM and the hardware companies would reach out to TrueNAS and shape. Maybe that’s a little idyllic but sitting at a stalemate doesn’t seem unacceptable. Many of the underlying technologies they use like zfs, Kubernetes, docker support ARM. If intel is building an ARM based chip, while the largest datacenters still primarily use ARM and it’s growing, and windows is starting to support ARM then I can only imagine it’s on their roadmap.

1 Like

Oh, and massive amounts of government (that is, taxpayer) funding.

Yep, money inspires.