Next Version of TrueNAS CORE

[adrianwi]

13.3 makes more sense, especially if it will be the last one

 

[Kris Moore]

This is one of the main advantages of FreeBSD over Linux, yet ironically it is not leveraged by TrueNAS Core.

But there are no ABI changes within one freebsd-stable train. These are maintenance releases. Why are you maintaining local modifications at all instead of deploying on a stock FreeBSD release?

Yep, this is also the reason why FreeBSD project is still maintaining the 13.x even though they already have 14.0. FreeBSD 14.0 has breaking ABI changes, the most notable one I've noticed being libssl.

Anybody who thinks you can deploy a production / enterprise grade appliance on a stock version of FreeBSD, ABI compat or no ABI compat clearly has never done this before. You also can't easily do that on Debian which is better supported and has superior ABI compat for binaries than FreeBSD does. The bottom line is that both of those platforms are designed to be generic servers and not specific to any particular appliance use-case.

When building an appliance, initial development is only 10-15% of the effort. 85% of the effort of making something enterprise grade is supporting it after the fact, which is where most projects fall flat on their face and end up being a buggy, unmaintainable mess in the long run. But I don't blame folks for thinking that there is a utopia of running on a stock version of some generic OS code. The cold reality is that it ends up being a quick path to sadness for the developers and users alike.

 

[bkw777]

What are the plugins good for anyway? I never looked at that because I use TrueNAS for, you know, a NAS.

Plugins were a way to try to provide a preconfigured jail for an application like a docker.

There are many such apps that would be a natural fit to host on your nas, like syncthing, or jellyfin, or immitch, or backuppc, all apps that act on the data on the nas, and don't tax the cpu or ram enough to even notice even on an old freenas-mini.

Well the apps and the jails is all fine and all, but in the end it's less grief to just use ordinary jails and install your apps in them the normal way rather than use the plugins.

The plugins used annoying configuration within the jail to make the jail simple to manage from the host by a few simple buttons like start/stop/update, but it just made the jail ultimately actually harder to manage by being weird inside, while not actually kept updated updated by the plugin authors so you did have to go mucking around in them manually or else live with old app software.

It's not the worst idea in principle, it just didn't produce enough benefit over a regular manual jail to be worth the cost of being so nonstandard.

Conversely even though the plugin system didn't really take off, the basic thing of running app jails on the nas, which is both 99% idle and also where all the data is, is valid for a home or small business system.

 

[mattsteg]

The plugins used annoying configuration within the jail to make the jail simple to manage from the host by a few simple buttons like start/stop/update, but it just made the jail ultimately actually harder to manage by being weird inside, while not actually kept updated updated by the plugin authors so you did have to go mucking around in them manually or else live with old app software.

It's not the worst idea in principle, it just didn't produce enough benefit over a regular manual jail to be worth the cost of being so nonstandard.

Plugins as 1-click to install and then at least the option to hand-off to the end-user to manage more like "regular" jails would have been a much better compromise in the long run. The update funkiness killed their viability, but 1-click installation and basic configuration was convenient. Maybe some bonus points for an update script.