Scyto

+ 1 for this and azure blob support (cold and warm, i already backup to azure using synology)

Integrated default reverse proxy for docker apps so all apps available on 443 inside home network. integrated DNS to support app name.internaldomain.com with reverse proxy (this puts all apps on 443) Lets encrypt SSL by default for all apps all user has to do is buy domain name and enter it above and all just works support for split horizon DNS consider running the DNS service for folks to allow tight integration (this shouldn't preclude use of other dns services) this is not me advocating for how to expose anything externally, that's not an argument i want to get into 😉

what would you like to know, i have it working..... 🙂 @Snrrlax

this already works 🙂 some limitations (need to manually specifiy the IPv4 address at each end, unclear if that perssists)

if you add thunderbolt.host_reset=false to the kernel command line then nvidia eGPU will works (i have tested this and confirmed this changes works with: sudo midclt call system.advanced.update '{"kernel_extra_options": "thunderbolt.host_reset=false"}' working (note a non-working system will show D3COLD errors in the output below, this is an issue on all recent LTS and mainline kernels on systems with USB4) truenas_admin@HexOS[~]$ sudo dmesg | grep nvidia [sudo] password for truenas_admin: [ 36.563182] audit: type=1400 audit(1733276217.971:3): apparmor="STATUS" operation="profile_load" profile="unconfined" name="nvidia_modprobe" pid=2363 comm="apparmor_parser" [ 36.563296] audit: type=1400 audit(1733276217.971:4): apparmor="STATUS" operation="profile_load" profile="unconfined" name="nvidia_modprobe//kmod" pid=2363 comm="apparmor_parser" [ 114.962824] nvidia-nvlink: Nvlink Core is being initialized, major device number 235 [ 114.968129] nvidia 0000:39:00.0: vgaarb: VGA decodes changed: olddecodes=io+mem,decodes=none:owns=none [ 115.122803] nvidia_uvm: module uses symbols nvUvmInterfaceDisableAccessCntr from proprietary module nvidia, inheriting taint. [ 115.148763] nvidia-uvm: Loaded the UVM driver, major device number 511. [ 115.156325] nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 550.135 Wed Nov 13 23:35:58 UTC 2024 [ 115.161201] [drm] [nvidia-drm] [GPU ID 0x00003900] Loading driver [ 115.161203] [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:39:00.0 on minor 1 [ 115.165157] [drm] [nvidia-drm] [GPU ID 0x00003900] Unloading driver [ 115.224654] nvidia-modeset: Unloading [ 115.273993] nvidia-uvm: Unloaded the UVM driver. [ 115.331235] nvidia-nvlink: Unregistered Nvlink Core, major device number 235 [ 123.378230] nvidia-nvlink: Nvlink Core is being initialized, major device number 236 [ 123.384055] nvidia 0000:39:00.0: vgaarb: VGA decodes changed: olddecodes=none,decodes=none:owns=none [ 123.507023] nvidia-modeset: Loading NVIDIA Kernel Mode Setting Driver for UNIX platforms 550.135 Wed Nov 13 23:35:58 UTC 2024 [ 123.509099] [drm] [nvidia-drm] [GPU ID 0x00003900] Loading driver [ 123.509100] [drm] Initialized nvidia-drm 0.0.0 20160202 for 0000:39:00.0 on minor 1

fair opinion, not one that is on the roadmap, the point here is for this to simplify truenas, at least at this time the team are committed to using truenas scale as is with ZFS. TBH there is zero point using truenas as a base if one is not going to use ZFS (for example see ZimaOS which uses normal mdraid and buildroot as its base). So while your request is fair its unlikely to be on that gets serviced by this project. But hey fingers crossed 😉

It supports nvidia GPUs for apps, 2080, 30xx and 40xxx all work fine. This is already in truenas.

Nice idea, especially optimizing the hexos curated apps to put different elements of apps into the two pools. For example immich storing photos on HDD but thumbnail and database data on SSD. Keep the idea going, it is a neat one!

If i were to guess it won't make any difference, they scale any cloud resources based on usage patterns of users in aggregate. Per signed-up customer scaling is not something they will have implemented as that drives costs in insane ways (don't ask me how i know, rofl)

Many folks seems to struggle with making this all work A default app that: puts all other apps by default on 443 for appname.local and for those that have DNS name on appname.mydomian.mytld bootstraps certs for both (self signed for .local and acme signed for mydomain) still leaves how to make acme DNS challenge 'easy or seamless' - but this sort of simplicity makes changing for OS easy (i guess you could run a subscription service for doing DNS name hosting, challenges, etc?) this default proxy should also be able to be turned off for those that wish to roll their own (say NPM or traefik) this proposal is making no statement about external connectivity as that discussion always turns into a poop-show of competing opinions.

I plan to use on my test rig: zimacube pro eGPU via thunderbolt (nvidia) 2 octane 905p 6 x 24 TB seagate drives 1 x 4 nvme switch card 1 x 2 nvme switch card 8 x nvme all working on latest truenas scale