You probably want to run all of that on bare metal in containers with Docker or Kubernetes.

Containers let you easily share resources between them, because they all share the same kernel. VMs are harder to share hardware resources with, as you're finding out.

I was not sure if I should run a LXC container for each docker, or have a single LXC with everything (exception xrdp / XFCE). I don't know what would be good practices..

LXC is a container. I don't think you would want to run Docker inside LXC. That's running a container inside a container. I'm a noob though.

Normally, you run one app per container, or one set of apps per container if they are closely related. You could run all the Plex suite apps inside a single LXC container and Windows alongside it in Proxmox. Or you could run each app inside their own LXC container.

Alternatively, you could run them all in individual Docker containers on bare metal Ubuntu, but not have the ability to install Windows or other OSes.

I run multiple lxc’s hosting portainer, there I can manage all my docker containers in 1 web ui. Its amazing. You can share files with samba or cifs in unprivileged containers (issue 1). I use that for plex because it’s easier to use the igpu in a lxc then a vm. In portainer you can bind mount your docker volumes to existing folders in your lxc and then proceed to share them with samba/cifs/nfs (issue 2). Note that nfs mounting in unprivileged containers is not (easily) doable. Portainer fixes issue 3. I can’t help with issue 4 but aren’t nics already hardware bound?

well, you said it yourself, get proxmox and take it forward from there.

Plex kernel is optimized with fixes specific to proxmox scenarios. For example it is currently impossible to do my setup https://gist.github.com/scyto/76e94832927a89d977ea989da157e9dc on Debian or Ubuntu without compiling your own kernel, which then would break ceph…. The proxmox team do a lot of due diligence on their kernel, patches etc. for your docker scenario consider using VMs as your docker hosts. Lightweight VMs don’t add a lot of overhead. Something like this https://gist.github.com/scyto/f4624361c4e8c3be2aad9b3f0073c7f9

I use proxmox with ceph and cephfs.

I have two ceph pools. One with ssd/nvme, and the other with slow rpm hdd.

The ceph pool with ssd holds my vhds for vms. I only install the os, on that drive...and any necessary programs..Docker, portainer, etc. Then I mount my two cephfs into each vm.

I don't use lxc...they can't be migrated between proxmox nodes. I just install Ubuntu server and Docker.

Since you mentioned a windows vm...cephfs can be mounted in Windows as well...no need to setup samba. No need to have data trapped in different vhds. Useful for mapping drives on laptops and desktops in your house too.

Cephfs is my central storage that all servers and containers have access too. If you want to have a NAS gui for users, just map your container to your existing files. I do this for nextcloud. I also run containers like sftpgo, filebrowser, syncthing, etc...all mapped to the two cephfs pools so I can expose and share single data pool via different methods and not have a single vm running samba that's a single point of failure.

Cephfs and proxmox support clustering, so if you need to grow or upgrade on the future. If you use proxmox for the second server, ceph and cephfs can extend the pool to the second server and moving vms is seamless because the storage is one big pool now. Run some vms on the new server and some on the old...with a single storage pool.

Ceph can grow while online, and is fault tolerant like raid and zfs.

Proxmox will also provide a web interface and console access to your VMs.

I group my Docker containers by function. So I'll have one Ubuntu Docker server for the Starr apps...then I have another with paperless, Nginx ubiquity controller, filebrowser, sftpgo...then another Ubuntu Docker for nextcloud because it spins up like 10 containers on its own. Another for Plex and jellyfin. Just trying to keep the chaos organized and because I have more than one server...I can move the one of the Docker servers to another proxmox node in my cluster to manually balance cpu and compute.

Then when it's time to upgrade, I'm not rebuilding everything...I'm just adding a small node with a small cost (small IT budget at home too 😂)

1. My media is on data-pool/media dataset but I was unable to share that with multiple LXC

I've never transitioned into proxmox from your specific setup but here's what you could do in proxmox: connect your pools to some kind of SATA drive controller, pass that controller through to a NAS virtual machine, share those drives out as NFS drives, and add those NFS drives as storage from the Proxmox web GUI. Then you can edit your LXC conf files to mount directories from the Proxmox host to the LXCs as needed.

2. All my docker configs are mapped as volumes on the host, and are stored on data-pools/apps. I would like to reuse them

I have successfully migrated docker environments from one host to another before simply by copy/pasting the folders and docker compose up -d. Then again sometimes it has not worked for me, but only one way to find out.

3. I was not sure if I should run a LXC container for each docker, or have a single LXC with everything (exception xrdp / XFCE). I don't know what would be good practices..

I don't know what good practice is but I have a couple LXCs with 12+ docker containers each and they work great for me 🤷‍♂️

Proxmox has a learning curve but I love it. It is so easy to spin up another VM to try out something new, and you don't have to interrupt the services that you already have working well because of how isolated everything can be from each other. Need to reboot your Linux environment? You no longer have to take down your entire server, just the part of your server that you're tinkering with.

Also the UI makes it so easy to manage backups, snapshots, you can create templates of prebuilt images so that you aren't starting from scratch each time, easily clone or migrate your images around to different nodes. It's very powerful.