It’s a pity that iX never seem to learn from their previous annoyances.
When they replaced Kubernetes with Docker, they could have built it to run both in parallel, but they chose not too, thus forcing any TrueCharts users to have to plan their upgrade much more carefully and in much more detail, documenting all their apps, and then run the TrueNAS upgrade and then immediately replace their TrueCharts apps with something else that might need a lot of experimentation or tweaking. This was perhaps 10x the amount of effort for these users than if there had been parallel running.
And now they have done the same with virtualisation. I have no idea whether it would have been technically possible to run old-style virtualisation alongside Incus or not, but since both use KVM just with different device emulation, I would guess that it would not only be possible but perhaps even easy.
But (yet again) iX chose: A) not to deliver parallel running to allow for easier migration; and B) to deliver a virtualisation technology that was NOT as comprehensive as that which it replaced.
If they did this then they could also switch to a one-a-year strategy for new versions (for when they finished parallel running and removed old technologies), and deliver new technologies as major point releases with bug fixes as minor point releases - and the overall impact would be far far smoother for both Enterprise and free users.
TBH I have given up hope that iX will ever learn this lesson. Of course there are a massive amount of plus points with TrueNAS which far far outweigh these negatives (which is why we are all still here), BUT there would be a lot lot less to gripe about if iX just learned to deliver new technology in parallel with the old whenever possible.
I think the reality is that there are very few paying customers that use TrueNAS for virtualisation, iX can therefore afford to continue to make breaking changes while they figure out the path forward for the product. Community edition meanwhile is clearly the vehicle for broad testing and feedback from enthusiasts - iX has been more or less explicit on this. They have also demonstrated through their actions that keeping that experience smooth across upgrades is not a priority.
So it’s up to everyone to decide if that works for them or not. Personally, as I think I mentioned before, I’m ok with that approach at the ”edge” of the product and new features like virtualization (which I don’t use anyway) - should it start to destabilise closer to the core of the actual NAS product though, putting data at risk, I’ll be moving off promptly.
Indeed. CORE is, or was, nearly equivalent to the Enterprise product, minus a few features which were of no relevance in home use, and thus very mature and stable.
SCALE/CE is now a perpetual beta, with only the storage part being stable, on a twice-a-year pace that is too fast to polish anything. Though if xx.04 is consistently used to introduce breaking changes which stabilise in xx.10, updating once a year could be an option.
I think 10x is an exaggeration. A contributing factor to why TrueCharts users had a difficult time was with how the TC-team had built things, hiding stuff away in hard to access places and then springing repository killing changes on the users with no warning what-so-ever.
What isn’t an exaggeration is that building and supporting two parallel apps layers would likely mean more than 2x the required developer time, since they would need to have a more complex set of code and UI handling two entirely different ways of doing the same things while not interfering with each other. I also question if it would be easier for end users since they would have two apps options mixing things up. I shiver thinking of the type of support requests we would have seen in the forums if they had gone that route.
At someone who didn’t use TC and had around 10 custom apps in 24.04, I migrated to 25.04 last week and it took me less than an hour to get 8 of them up and running using custom YAML, an hour to get Nginx Proxy Manager set up (I used the LinuxServer nginxreverseproxy container previously) and another hour to get Jellyfin back up, because of reasons. I will also point out that I did not have to swap back in 2024, I had plenty of time to settle on a migration plan since iX informed us of their coming changes over a year ago.
I have yet to try the new Incus parts so I can’t say if that swap was premature.
I’m in the situation of running a single TrueNAS SCALE (sorry, Community) box as a home server with VMs that I use daily. So not “mission critical” in the sense that there’s a business depending on it, but definitely in the sense that there will be weeping and gnashing of teeth if the systems I and my family use are scuppered by a ropey upgrade.
So we have 25.04 labelled as not being for general use and virtualisation being labelled as experimental in the release notes, yet Fangtooth is the version advertised for download from the Community > TrueNAS Community Edition, where “KVM VM Support” is listed as product headline 3. Maybe the marketing department weren’t invited to the same meetings as the product team?
What I’d like to know is whether virtualisation is being planned to move out of “experimental” at any time during 25.04’s lifecycle, or if maturity is only expected at 25.10. If it’s the latter, my next question given last year’s Kubernetes to Docker handbrake turn is, what feature will be broken in 25.10??
I understand the premise (even if iX haven’t said it out loud) that TrueNAS Community users are often early adopters who can help them test out changes before shipping to their paying enterprise customers. But at this point isn’t it fair to ask for some kind of roadmap for what lies ahead, so we can at least try and live our lives without the 6-monthly experience of having to deal with the fallout of what ever whim has led iX to change out another feature? Please?!
EDIT: Just to belabour the point, on the Community Edition download page, 25.04 is labelled with a big green “STABLE” button, with 24.10.2.1 labelled with an off-putting grey “LEGACY” button. So is 25.04 for developers, testers and early adopters as per the status page, or is it the green-button happy times for everyone of the download page?
That would reason enough to never upgrade to a .0 release but wait for .2 or .3, depending on feedback from earlier adopters here.
Then you’re in trouble, indeed…
Maybe you could try 25.04 virtualised somewhere to see if you can successfully recreate your VMs under the new system.
Or hold on 24.10 until 25.10(.x) is out.
Good point. That is at best excessively optimistic, and at worst misleading, for a .0 release with a breaking change in functionality and acknowledged work-in-progress around the changes.
You can just do it like me and stay on 24.10 and wait for 25.10.
I myself depend on Windows VM running smoothly so I need for example VM snapshots being supported. So I will just wait until they are.
My Truenas works perfectly fine on 24.10 so I have no need to upgrade before its ready.
Well, for new users the 25.04 is stable and perfectly usable because new users dont have existing VMs they depends on. So there is no problem.
I guess it depends on meaning.
Maybe 25.04 can be both stable and also only for early adopters?
Or at worst its just simple mislabeling and they should wait until its labeled as “for general user” until they put it under the big green stable button.
My point is that only thing in the “experimental” stage is Instances section. So everything else is stable and good to use, just this one part is not.
And for new user thats not that bad. They can just try it and if its missing some features they need they can just not use it.
@Foxtrot314 granted, but running HAOS in a VM is really not that uncommon among enthusiasts. Virtualisation is almost always better than adding Raspberry Pi after Raspberry Pi for your applications
And for example a close friend who is not a “serious home labber” at all like perhaps myself - he absolutely depends on Home Assistant running. Lives in a house (a german one built from bricks, mind you) and his entire energy and heating managament is done in HA.
Additionally he runs some applications in jails - on TN CORE, among them Paperless-NGX thanks to the work of @victor. That’s the document archive for their family for crying out loud.
And yes, there’s a backup to another TrueNAS in a completely different location. And local snapshots. And replication from the SSD “jail and VM” pool to the HDD “storage” pool etc.
So a couple of weeks ago I naturally suggested he move to SCALE. Paperless and other stuff in an app. Home Assistant in a VM. Not quite as flexible as the CORE system in terms of networking but we can put the apps into the family VLAN and the HA VM into the “chinese IoT stuff” VLAN. Two physical network interfaces, easy peasy.
And I think that is a typical use case that is perfectly matched by what TrueNAS has to offer.
Guess my reaction learning they completely revamped VMs again. I’ll recommend to him to stay on 24.10 and just relax. And after all it’s dead easy to move a complete VM anywhere else. Free ESXi if you must
I just hope this support nightmare will end some day and he will be able to just click on some update button and not need my help restructuring everything every single time.
Personally I will continue to run CE (25.04 currently) for Docker based workloads but stay on FreeBSD for everything else. For reasons that go far beyond OS of choice zealotry.
Likely just different forces at play, and different people that wrote the “software status” page vs those who designed the download page. There’s clearly a discrepancy when the status page has 24.10 for general use and 25.04 for “early adopters”, whereas the download page - which will be as far as most users go - screams green STABLE in capital letters for 25.04 and anything else “legacy”, without further clarification. Who is really going to press the “legacy” button? And what percentage of new users are going to click their way deeper into the software status page before deciding?
Of course iX knows all this. And clearly they want as many users as possible on latest versions or even RCs and betas, because it provides testing coverage and feedback on the product. If they get users to migrate more quickly it also means fewer previous versions to maintain in parallel. So they will continue to be conflicted between maintaining a stable enough product for their community crowd, vs getting users to test and evaluate changes and new features. I think these contradicting pages on the web site just highlights this built-in tension, and therefore comes across as misleading or possibly even disingenuous to those who have a deeper understanding and interest in the product.
I don’t think that is true unfortunately. But let’s see how many and what type of fixes make it into 25.04.1.
Ouch! Any specific concern, or is it more of a “general feeling”?
I personally absolutely expect the ZFS NAS part to be stable—if not SCALE would just not be fit for purpose. I also expect the Docker app part to be stable and good to use now that the breaking change is passed—and upgrading to 25.04 implies that one is ready for June 1st.
It seems that even legacy nspawn sandboxes (jailmaker) survive an upgrade to 25.04. But (the APIs for) the new LXC sandboxes and VMs are not stable yet.
Somewhere in between tbh. The product has gone from “we release when ready” (Debian-style) to a rolling 6 month release cycle. So consider typical project/product management - quality, features and speed/time-to-market typically triangulate. Add cost if you like as a 4th. Meaning if you fix one, one or several of the remaining have to move, all else equal. iX has now fixed speed/time-to-market with their 6mth cadence. Between the remaining, they chose to prioritise features over quality in 25.04, at least in the area of virtualisation. Meaning they released something that was in alpha-status as part of their “stable” product. That looks to me like a clear change in an attitude vs before so of course begs the question if we’ll see similar trade-offs in other areas too, or if we have already in 25.04.
Add to this what I perceive - rightly or wrongly - a relatively large number of reported issues on this forum not just in virtualisation but also around middlewared, kernel drivers and other areas.
YMMV but I’ll wait a few point releases before upgrading.
Fair points. I’ll count the “final release of CORE 13.3-U1.2 and we won’t roll in the one-liner U2 fix that was already on the shelf and ready for months” as another symptom of the change in attitude.
The system creates a bunch of network interfaces automatically named like “veth*” or “br-*”. As I understand this can be the consequence of using Docker (I do).
I have literally one Debian VM with br0 and br1 attached to it. No MACVLAN, nothing complicated.
br0 is a bridge with enp10s0 for Internet access, while br1 is a bridge without any physical interface for private “host-only network” access. (I created them on the “Network” control panel in TN).
