One Year of TrueNAS SCALE 25.10 Goldeye in Production: Our Honest Review Before the 27 Upgrade

On 28 October 2025, iXsystems released TrueNAS SCALE 25.10 Goldeye. Almost exactly one year later — and two months before the successor TrueNAS 27 reaches GA in early December 2026 — is the right moment for an honest interim review. Not the marketing version, but the one that comes out of twenty productive installations at DATAZONE customers.

This article is a hands-on review: what actually ran stably over the months, where we had to intervene in operations, how enterprise support and the community edition differ in practice, and which questions are now on the table before the upgrade to TrueNAS 27. As an authorised iXsystems partner in Bavaria, we have access to the same support channels as our enterprise customers — our take is correspondingly honest, but constructive.

The scope: what we have seen in one year of Goldeye

Our installed base of TrueNAS 25.10 ranges from small Mini R systems for home-office setups to redundant M-Series clusters with NVMe-oF. In between there are classic R-Series boxes at mid-sized manufacturers, F-Series flash nodes for virtualisation back-ends and a few community-edition installs on custom hardware. The mix is intentional: we see Goldeye not just in one niche but across the breadth of what iXsystems designed the platform for.

Most of these systems either went into production directly on Goldeye or were migrated from 24.10 Electric Eel during the year. That means we loaded both the greenfield and the upgrade path. None of the systems had data loss during the year. Two had hardware-induced outages, one hit a configuration pitfall that cost us two Monday mornings. More on that below.

What really ran smoothly over a year of operation

Let us start with the positive, because that accounts for the bulk of the observation: Goldeye is a remarkably stable release across the board. The core functions — ZFS, SMB, NFS, iSCSI, snapshots, replication — run quietly. Quiet is exactly what you want in a storage operating system. Of the roughly twenty systems, we had exactly two unplanned reboots during the year, both caused by hardware (once memory, once a motherboard fault). The operating system itself never produced a kernel panic.

NVMe over Fabrics in practice

NVMe-oF over TCP in the community edition and with RDMA in enterprise was the headline feature of 2025. In our practice, two customers used it actively — once as a Proxmox cluster back-end over 100 GbE, once for a VMware setup over 25 GbE. Both installations have been running stably for over ten months. We deliberately keep performance figures to ourselves, because we do not publish synthetic benchmarks we cannot reproduce in every customer environment. The short version: for the intended purpose it works, and the configuration overhead was manageable.

Replication between sites

Encrypted dataset replication between a primary and a backup location is in use at roughly two thirds of our customers. That worked pleasantly quietly throughout the year. The few issues we had were network-related — dropped VPN tunnels, DNS changes on the customer side — not TrueNAS-side. Snapshot policies with tiered retention times work as documented, and recovery tests consistently gave us a positive outcome.

The new dashboard and UI

The reworked interface in Goldeye has proven itself in daily use. Admins who had previously worked with 24.10 were up to speed within a few days. Newcomers — and we had a number of those during the year, especially customers migrating from Synology or QNAP — find the key functions far faster than in earlier versions. That is not a small thing: UI learning curve is an underestimated time sink in every project.

Where we had to intervene: the honest pitfalls

Now the other side. Goldeye was not flawless, and in two or three points we would have appreciated a faster reaction from Santa Clara.

The apps catalogue in the first months

The new apps catalogue with the reworked container back-end was noticeably rough in the first two to three months after release. Individual apps did not update cleanly, we had to intervene manually on a Nextcloud setup after a minor update, and the logs were temporarily harder to read than usual. With the dot releases 25.10.1, 25.10.2 and by 25.10.4 at the latest, this calmed down significantly. Anyone setting up a fresh system today does not notice it any more. Anyone who went in six months earlier spent a few evenings on it. Lesson for us: for productive app workloads we wait for the second or third dot release on a major, and we use community systems in the first weeks specifically to collect experience.

A configuration topic that caught us twice

At two customers we independently saw the same phenomenon: a ZFS dataset with very specific inheritance settings for SMB ACLs behaved differently after an update than previously documented. Specifically, certain permissions no longer inherited cleanly down to child datasets, which creates confusion in grown share structures. This was not a bug in the classic sense, but a behavioural change that was not sufficiently emphasised in the release notes. We resolved it cleanly with the iXsystems support — the response time in enterprise support was under two hours both times — but it is the kind of pitfall you want to avoid when release notes are documented thoroughly.

Secure Boot and older VMs

Secure Boot for virtual machines is a good idea, but with older Linux VMs that previously ran without it, the migration path was bumpier than necessary. For Windows guests it was unobtrusive, for CentOS 7 legacy workloads at a manufacturing customer we had to build a specific boot configuration. Nothing an admin cannot solve, but a time sink when the target group is heterogeneous.

Enterprise support versus community: what is the difference

A question we hear most often in consulting conversations. In our practice it looks like this:

Enterprise support with iXsystems means primarily two things: guaranteed response times for severity-classified tickets, and access to hotfixes before they are published. The latter was relevant three times during the year — in each case when a customer hit an issue for which a fix version already existed in the pipeline. Response times in our experience were regularly under two hours for severity 2 and 3, and under one hour for severity 1.

Community-edition users get the same code and the same features except for NVMe-oF RDMA and VM HA failover. What is missing is the ticket system. For home labs, small non-profit setups and development environments that is entirely fine. For productive SMB environments with revenue impact we recommend enterprise support — not out of dogma, but because the math works out in a single real incident. How we support our customers is described here.

Partner access and what it means in practice

As an authorised iXsystems partner in Germany, we have an intermediate layer: we can open tickets on behalf of customers, we receive pre-release information and we regularly sit in technical calls with the engineering team. That does not mean we work magic — but we often reduce the number of follow-up questions between customer and vendor, because we know the setup in detail and pre-sort the problem for support.

Hardware observations from one year of Goldeye

On hardware briefly and factually: the current Mini R, F-Series, R-Series and M-Series M30/M40 systems run with Goldeye without incident. The M50 and M60 are now EOL and should no longer be chosen for new projects — existing devices continue to receive updates, but anyone planning something new picks M30 or M40.

For concrete hardware selection, our TrueNAS configurator is the fastest way to work through pool sizes, network interfaces and replication topologies in one pass. We do not quote fixed prices there because component prices move on a quarterly basis; instead interested parties receive a current price range or an individual quote.

Preparing for the 27 upgrade: what makes sense now

TrueNAS 27 reaches GA in early December 2026. For most of our Goldeye installations that does not mean an immediate upgrade. Our internal recommendation for productive systems: observe 27.0, test 27.0.1 or 27.0.2, plan 27.0.3 for production — realistically Q1 to Q2 2027. For community-edition setups and test systems the upgrade can come earlier, especially if one of the 27 features is specifically needed.

Checklist before the switch

What is always worth doing in preparation for the 27 upgrade — regardless of when it actually happens:

• Export the configuration: System → General → Save Config. This is elementary and gets forgotten regularly.
• Check pool status: all pools healthy, clean resilver history, up-to-date scrub cycles. You never start an upgrade on a degraded pool.
• Document the app catalogue: which apps run in which version, with which persistent volumes. That avoids searching if an app has to be recreated after the upgrade.
• Validate replication targets: if a second site is to serve as a fallback layer, it must be current and freshly replicated.
• Check hardware compatibility: the TrueNAS 27 release notes will define which older controllers and network cards are still supported. For existing hardware this is the central check.

For enterprise customers we offer this as a structured upgrade project — including a test run on a mirrored system before the productive device is touched. More on IT service around TrueNAS can be found here.

Conclusion after one year: solid base with small scars

TrueNAS SCALE 25.10 Goldeye was and is a good release. It is not flawless, but the flaws we saw were manageable, documentable and solvable with vendor support. The core functions — ZFS integrity, snapshots, replication, SMB/NFS/iSCSI — delivered across twenty installations for a year without us having to seriously worry. The new features — NVMe-oF, Secure Boot for VMs, reworked dashboard — are stable in the main as well, with the apps catalogue being the only larger initial weakness.

For the road to TrueNAS 27 this means: Goldeye is a solid starting point. Anyone who is on 25.10.6 today and has their infrastructure documented cleanly can approach the 27 upgrade calmly once the second or third dot release is out. For new installations until the end of the year we continue to recommend Goldeye — you do not build a productive environment on a release that has not been GA for two weeks.

If you need a well-founded second opinion for your setup — whether on upgrade path, hardware selection or support model — get in touch with us. We have seen the cases where things went smoothly, and the ones where we pulled the vendor in late in the evening. Both feed into our advice.

FAQ

Is it still worth upgrading from TrueNAS 24.10 Electric Eel to 25.10 Goldeye now?

Yes, in principle. Goldeye in the current dot version 25.10.6 brings along the benefits of a year of bug fixing. Anyone still on 24.10 should plan the move to 25.10.6 before TrueNAS 27 goes GA — so the next upgrade cycle starts from a cleanly documented base. For enterprise customers we support this step as a structured project.

Which TrueNAS hardware is sensible for a new deployment with Goldeye?

For smaller setups and home-office scenarios the Mini R line, for mid-sized SMB environments the R-Series or F-Series when flash is needed, for high-end projects the M30 or M40. The M50 and M60 are EOL and no longer intended for new projects — existing devices continue to receive updates, but product care has ended. Concrete configurations can be worked through in the TrueNAS configurator.

Was the apps catalogue in Goldeye really a problem?

In the first two to three months after release: yes, in parts. Individual apps did not update cleanly, the logs were less readable than usual, and in more complex setups like Nextcloud we had to intervene manually several times. By 25.10.4 at the latest this was behind us. Anyone starting fresh today does not notice it.

When should one upgrade to TrueNAS 27?

For productive systems we recommend observing 27.0 and waiting for 27.0.2 or 27.0.3 — realistically Q1 to Q2 2027. For community-edition and test systems the upgrade can happen earlier, especially if a concrete 27 feature is needed. More background on this is in our note on the TrueNAS 27 release.

Is enterprise support from iXsystems worth it if you only run a small environment?

That depends on the revenue relevance of the storage, not on the size of the system. A 20-terabyte system in a law firm, where one day of downtime immediately means lost revenue, justifies enterprise support just as much as a 300-terabyte manufacturing setup. For pure test systems and home labs the community edition is absolutely sufficient.

What does DATAZONE as an iXsystems partner actually do differently than a pure hardware reseller?

We sell the same hardware as other authorised partners — but we also set it up, migrate existing data, build replication targets and support the systems in ongoing operation. On top of that we have direct access to engineering and support channels at iXsystems, which in real incidents makes the difference between hours and days. More on this on our TrueNAS support page.