Is TrueNAS 26 Worth Running While It Is Still in Beta?

Start with the part the feature announcements bury. As of August 2026, TrueNAS 26 “Halfmoon” has shipped exactly two builds — BETA.1 on 7 April and BETA.2 on 17 June — and iXsystems has announced no GA date at all. Everything below is therefore a question about a pre-release, which is the answer to “should I upgrade” before any feature enters the conversation.

What makes the release interesting anyway is not the feature list. It is a deliberate trade: iXsystems gave up release frequency for an annual cadence with point releases, betting one carefully-baked major per year beats two rushed ones. That bet either buys stability or concentrates a year of risk into a single event. Halfmoon is where we find out which.

TrueNAS 26 Halfmoon Is Still at Beta.2, and That Is the Whole Answer

Two betas exist. BETA.1 arrived 7 April 2026 on kernel 6.18.13; BETA.2 followed on 17 June 2026 on kernel 6.18.23. Both carry OpenZFS 2.4.1. Between them the kernel moved ten patch versions inside the same LTS line — the shape of a release still absorbing fixes, not one settling down.

A beta designation is a support statement, not a quality guess. iXsystems has not committed to standing behind this code on your production array, so any Enterprise support conversation about a failure on 26 starts from a worse place than the same conversation about 25.10. The predecessor, TrueNAS 25.10 “Goldeye,” has been generally available since 28 October 2025 with maintenance releases landing on it ever since. Nine months of field hardening against two beta drops is the actual comparison, and no feature in Halfmoon closes it.

So the upgrade question splits cleanly in two. On a lab box or a spare pool you can reinstall without a change ticket, running Halfmoon now is useful and cheap. On anything holding data you would miss, the answer is not “no forever” — it is “not until there is a GA build with at least one point release behind it.”

Five Ways TrueNAS 26 Pulls Ahead of 25.10 Goldeye

The feature set is substantial, and most of it targets the two things Goldeye left half-finished: containers and multiprotocol file serving. Here is what actually moves.

TrueNAS 26 Versus 25.10 Goldeye, Line by Line

The blank OpenZFS cell is deliberate: the 25.10 version notes do not state a version, and inventing one to fill the table is how a guess quietly becomes the internet’s answer.

The REST API Is Gone, and Your Automation Will Break

Here is the change nobody leads with. The TrueNAS 26 version notes state that the REST API — deprecated back in 25.04 — is removed in 26 and fully replaced by the versioned JSON-RPC 2.0 WebSocket API. Not soft-deprecated. Not behind a flag. Removed.

If you have ever pointed a cron job, an Ansible task, a monitoring check, or a five-line shell script at /api/v2.0, that code stops working the moment the system boots 26 — and it will not fail loudly, because the upgrade never scans your scripts. It fails later, quietly, in whatever prunes your snapshots or exports your capacity metrics.

Upgrading the Pool Is the One-Way Door, Not the Install

Before you touch a beta, understand that installing the release and upgrading the pool are two separate acts with very different consequences.

Installing 26 is reversible. The previous release stays in its boot environment, so if Halfmoon misbehaves you reboot into 25.10 and carry on — nothing on disk has changed format. That is what makes beta testing on real hardware survivable at all.

Upgrading the pool is not. That step writes new OpenZFS feature flags into the pool itself, and an older release no longer knows how to read what it finds; the rollback you had five minutes earlier is gone. The discipline is simple: run 26, use 26, benchmark 26 — and leave the pool version alone until 26 is a release you would have installed anyway.

Why CORE Holdouts Should Care About Full LXC Support

If you are still on TrueNAS CORE in 2026, it is almost certainly because of jails. FreeBSD jails held the services that made the array useful — a media server, a DNS resolver, a reverse proxy, some homegrown thing nobody has re-packaged — and the Linux-based line had no equivalent to move them into. So you stayed.

That is the gap 26 closes. Full LXC support, with Enterprise HA failover and GPU passthrough, is the first landing zone where a jail-shaped workload has a genuinely equivalent home: a persistent, lightweight Linux container you administer like a small machine rather than repackaging as an app. Nothing converts a jail for you — it is not a migration tool — but it is the first time the destination has been stable enough to plan against.

So LXC in 26 is the reason a CORE holdout should start planning the move — not the reason to run the beta this quarter. Build the container definitions on a test box, confirm your services behave under LXC, and be ready to execute when 26 goes GA. The planning has no downside; doing it on your production array during a beta does.

What the Annual Release Cadence Actually Buys You

Here is the thesis. TrueNAS 26 moves to an annual cadence with 26.x point releases, down from roughly two majors a year. iXsystems traded release frequency for confidence, and that trade only pays off under one condition: the point releases have to be genuinely boring.

Slower majors are usually good for storage admins. Every major version is a risk event — a kernel jump, a ZFS jump, an API removal — and halving them per year halves the maintenance windows you spend proving nothing broke. Each release also gets more soak time before the next pulls attention away.

The counter-argument is concentration. At two releases a year, a rough one was survivable because a better one was six months out. On an annual cycle a rough major is your platform for twelve months, and the pressure to cram a year of work into one build is exactly what produces a rough major. That risk is not hypothetical for infrastructure vendors — the VMware licensing upheaval showed how fast one vendor decision resets a team’s roadmap.

Which way it lands is measurable: if the 26.x point releases arrive on time and fix things without adding features, the trade worked. If the first one carries new functionality, the cadence changed on paper only.

Who Should Install TrueNAS 26 Today, and Who Should Not

Install it if you run a lab, a test pool, or a second array whose data lives somewhere else too. The kernel jump and the API removal both want real-hardware testing, and the people who file good beta bugs are why the GA build ends up boring. Just do not upgrade the pool, and do not let the test box quietly become something people depend on.

Wait if the array holds production data, backs anything up, or is half of an Enterprise HA pair. There is no EOL pressure on 25.10 and no GA build to move to, so the whole cost of moving early is yours. If this array is a copy in your backup rule, it is not a beta candidate.