Is anyone still running the jailmaker script that creates systemd-nspawn containers and could confirm this survives an upgrade to 25.04?
thank you!
1 Like
I’m still running my nspawn-jail for my docker and can confirm it’s still working on 25.04-RC and the .0 release from last week.
I’m still waiting for some features to be available in incus before migrating away from jailmaker.
I have already played around with incus on a test system and got my stack deployed there, apart from some trouble with port 53 for blocky and nvidia support for jellyfin.
4 Likes
Thank you for the quick reply!
1 Like
I am also still running with jailmaker
As @LarsR indicates I am also planning a move to an Incus “Jail” which is currently up, running and ready to go and will (should) be a relatively simple move the stacks over process but the “Experimental” nature of Instances plus no way to back them up even with snapshots etc make me reluctant to make the final move.
I understand that IX are working on the Instances system and a fuller approach will be available in the next release
I, unlike @LarsR am not concerned with nvidia support inside the instance as I run Plex using an App
Thank you too! Good feedback.
So is your plan then to create an LXC container and run your docker containers in there instead of systemd-nspawn?
I currently only have a single “jailmaker” environment which runs docker with three containers. None of these require nvidia support.
I am reluctant to use the app system given all the changes to that system over the recent versions.
yes, i’ll be migrating my 29 docker container from jailmaker to an incus lxc instance as soon as nvidia passthrough is properly supported.
For the port 53 issue i’ve alread found out that incus uses that port itself for dns and you have to remap that port to something else
As for nvdidia, it’s already supported to passthrough a gpu, but if you try to manually install the nvidia driver, you get a kernel miss-match error because the version truenas uses is newer then the driver that’s available from the package managers. Trying to manually install the same driver as truenas uses also throws an error (can’t remember what though).
That’s why i also created this feature request
I share the same reasons for sticking with Jailmaker. It works in 25.04 for me. Once 25.10 comes out and the instances feature matures I’ll be migrating from Jailmaker over to incus.
Nvidia works fine in Incus aka Instances. There are scripts that will set you up with docker in incus as well. See here.
Side note, I moved on from jailmaker months ago.
2 Likes
Not if you deploy it via Truenas gui, which is to be expected what the majority of users will use.
Yeah, jailmaker and incus are synonymous right now. Both have to configured with scripts. But at least with Incus, once it’s built, which takes about 90 seconds, it’s fully managed in the UI.
1 Like
That may be true, but there are users that are either not technical enough to configure it via script or want to use it the “truenas” way, which has always been to configure it using the gui and not shell commands (apart from maybe api calls). And these users are guaranteed to hit a wall when trying to use nvidia drivers inside an incus lxc.
I have just upgraded and can confirm my jailmaker works as expected.
Next step then is to test incus and see if I can move over to it.
1 Like
I don’t know why you need to install the Nvidia drivers in the container. That’s not needed.
Ok, then maybe i’m really stupid.
I’ve deployed a debian bookworm and trixie lxc from the truenas gui, ticked the checkbox to passthrough my nvidia T400 to the lxc and started the instance.
Trying to confirm the gpu get’s detected via nvidia-smi results in “bash nvidia-smi not found”.
The truenas gui does not pass through the host nvidia configuration. So it’s basically useless to have that gpu passthrough tickbox.
And without any further knowledge of incus and the documentation , what will you do? Try to install the drivers via package manager inside the lxc, which results in a kernel error because the driver truenas uses is newer then the one you can install via apt.
You’re not stupid.
You need to enable the Nvidia configs. That’s it. See the config I posted in the OP on the other thread. There are incus workarounds in my cloud init script as well.
You also need to run a job on boot to execute nvidia-smi.
Read through the OP as well.
But that’s the point, those are manual workarounds that ppl who are not familiar with incus simply do not know. There’s no mentioning of additional config options that are not exposed in the truenas incus wizard, no mentioning of the incus documentation.
If you’re a new user, have 0 experience with incus, it is pretty frustatring, and i really don’t think it’s feasable to have new users learn incus config options that really should be exposed by truenas.
There’s nothing manual to do, it’s in the configs, fully automated. I did all that already.
Only thing manual is creating a preinit job to initialize the driver.