TrueNAS CORE 13.3 has a new sustaining release available for download. This update solves a critical OpenZFS issue.
Error with device removal and block pointers remap with cloned blocks NAS-133555.
13.3 Final Release
This marks the final release for the TrueNAS CORE 13.3 software train. We extend our heartfelt thanks to all our community users who have journeyed with us throughout the life cycle of TrueNAS CORE 13.3. As we close this train, we invite you to explore our newest TrueNAS Community Edition solutions. TrueNAS 25.04 (Fangtooth) brings improvements to Apps and OpenZFS for both Community and Enterprise users, and is the recommended migration path for current 13.3 installations.
If any security or data integrity issues do arise, we will notify the Community of these. The expected resolution will be in the TrueNAS Community Edition.
Wait. They didn’t fix the bugs with the scrub schedules and boot-pool scrubs?
The bugs were known and reported back in September 2024, ever since they were introduced in 13.3-RELEASE.
I’m getting mixed messages.
This bug and this bug are reported as “Done” and “Fixed for version 13.3-U2”.
But the releases notes for 13.3-U1.2 says that this is the last update for Core.
Huh?
Am I to understand that iXsystems is saying: “Yes, we fixed the bugs that were reported in September 2024. The fixes will be included in TrueNAS Core 13.3-U2. Oh, and also… 13.3-U1.2 is the last release. ”
Yea, we’ve opted to make this the final release, U2 is no longer planned. The numbers just don’t justify it in any way, even this update was debatable. Users running 13.3 are <0.007% of total fleet. By way of comparison there are 3x as many users still running FreeNAS 11.3 which means the risk of running on 13.3 is far higher, just due to the lack of any real serious test and soak time in field.
Recommend folks find ways to migrate in near future, but ultimately we didn’t want to leave this ZFS-level corruption bug standing, even with the very low usage numbers.
Fair. But even so, this was way lower than my lowest of predictions. I figured there were enough Jails / FreeBSD diehards to be worth a paltry 4-5% number at minimum.
Its purely semantics, but done in this Jira context means that code was checked into a GitHub repo, but is untested. We’re not expecting to do any further test cycles or bring those changes into a future test sprint to prep for release.
However, folks can go grab those commits if desired and apply manually to test if it fixes the issues they are intended to. Since there will be no more updates of this train, you won’t have to worry about it regressing on a future update.
Untested for whom? The last holdovers on Core 13.3 are willing to take a “risk” for two fixed bugs to be included without testing. Especially since these bugs were introduced with 13.3.
I’m confused by the rationale. This isn’t for enterprise, nor does it tarnish SCALE. As a matter of fact, where was the “testing” in 13.3 when these news bugs snuck in?
Its worth understanding the economics of a minor release… e.g. 13.3.1 or 24.10.2
We’ve grown and invested significantly in QA over the last few years. Quality has improved dramatically, but the cost is that it takes about 3 weeks of over a dozen people in a lab worth a few $M. So each minor release costs close to $100K to get out the door.
For electric eel… that’s less than $1 per user
For 13.3 … that is more than $30 per user
These costs do not include developers time… just QA
Another fair point, I guess. But what would have been the incremental cost of rolling up everything in the final update, OpenZFS and the already fixed bugs?
Or, to put it otherwise, the cost to introduce the OpenZFS fix in the U2 branch rather than U1?
This hotfix did not have a full QA cycle because the same fix has been tested in other releases and is a known good patch. We assessed that this was best approach… rather than leaving this bug unfixed.
Highly, highly recommend you don’t, and move over to CE (25.04 and later) instead. It completely revamps VMs, and you don’t want to set your VMs up on SCALE and then need to redo them again for CE.
CE 25.10 is the first release that is meant to have great UX for the new Incus VM functionality, Enterprise ready. If you’re waiting, may as well wait for 25.10 and then migrate to that.
Good news is it’s not scary. With a config backup (including the secret!!) you’re back up and running on NAS functionality in no time. VMs are pretty easy to bring over, just make a new VM with the existing zvol; jails need to be redone entirely as Apps. Make sure your jails have their data in a separate dataset so that this is easy - if they don’t, make that change beforehand. Document every jail, what it does, where it keeps its data, how its networking works - and then yes while moving that to Apps takes time, it can be done bit by bit.
Wait. They didn’t fix the bugs with the scrub schedules and boot-pool scrubs?
”
