
When your media files are spread across multiple NAS boxes—some on your local LAN, others at a remote site—you don’t need to migrate terabytes of data just to stream from one place. Here is how I unified media from three separate NAS servers (nas1, nas2, nas3) onto a fourth node (nas4) running Jellyfin in rootless Podman over Tailscale.
The Topology

nas1&nas2: Remote site storage nodes, connected over the Tailscale mesh.nas3: Local LAN storage node (<lan-ip>), also hosting the homelab portal (homelab_portal).nas4: Dedicated compute node running Fedora (4-core/8-thread CPU, 32 GB RAM) — acting as the Jellyfin streaming host.
1. Deploying Jellyfin with Rootless Podman Quadlets
Instead of installing native RPMs or running Docker as root, I deployed Jellyfin as a rootless Podman Quadlet under ~/.config/containers/systemd/jellyfin.container on nas4:
[Unit]
Description=Jellyfin Media Server
Wants=sshfs-nas@nas1.service sshfs-nas@nas2.service sshfs-nas@nas3.service
After=network-online.target sshfs-nas@nas1.service sshfs-nas@nas2.service sshfs-nas@nas3.service
[Container]
Image=docker.io/jellyfin/jellyfin:latest
ContainerName=jellyfin
SecurityLabelDisable=true
PublishPort=8096:8096
Volume=%h/jellyfin/config:/config:Z
Volume=%h/jellyfin/cache:/cache:Z
Volume=%h/mnt:/media:ro,rslave
AutoUpdate=registry
[Service]
Restart=always
[Install]
WantedBy=default.target
Two important details here:
1. :ro,rslave on the /media mount ensures that FUSE/SSHFS mounts added on the host under ~/mnt/ propagate into the running container, and SecurityLabelDisable=true prevents SELinux from blocking access if a FUSE mountpoint temporarily drops or has a user_home_t label.
2. sudo loginctl enable-linger <user> keeps the user systemd service alive after logout or reboot.
2. Linking nas1, nas2, and nas3 via sshfs
With fuse-sshfs installed on nas4 and SSH keys authorized on nas1, nas2, and nas3, each remote share mounts cleanly into ~/mnt/:
OPTS="-o StrictHostKeyChecking=accept-new,reconnect,ServerAliveInterval=15,ServerAliveCountMax=3,idmap=user"
sshfs $OPTS <user>@<nas1-tailscale-ip>:/media/storage ~/mnt/nas1
sshfs $OPTS <user>@<nas2-tailscale-ip>:/media/storage ~/mnt/nas2
sshfs $OPTS <user>@<nas3-tailscale-ip>:/media/raid ~/mnt/nas3
Making sshfs Persistent Across Reboots (sshfs-nas@.service)
Manual sshfs mounts disappear on reboot. To make them self-recovering, I wrapped sshfs -f (foreground mode) in a parameterized systemd --user template (~/.config/systemd/user/sshfs-nas@.service) with Restart=always and automatic stale-mount cleanup (fusermount3 -uz) before starting:
[Unit]
Description=Persistent SSHFS Mount for %i
After=network-online.target
[Service]
Type=simple
ExecStartPre=/usr/bin/bash -c 'mkdir -p %h/mnt/%i && (mountpoint -q %h/mnt/%i && fusermount3 -uz %h/mnt/%i || true)'
ExecStart=%h/.local/bin/mount_nas_sshfs.sh %i
ExecStopPost=-/usr/bin/fusermount3 -uz %h/mnt/%i
Restart=always
RestartSec=10
[Install]
WantedBy=default.target
With systemctl --user enable --now sshfs-nas@nas1 sshfs-nas@nas2 sshfs-nas@nas3, all three shares mount automatically at boot before the jellyfin Quadlet starts.
Gotchas We Hit Along the Way
- Directory permissions on
nas2:/media/storageonnas2was711(drwx--x--x root:root), which causedPermission deniedwhen listing/media/nas2inside Jellyfin. A quicksudo chmod 755 /media/storageonnas2fixed it immediately. - Disable real-time library monitoring:
inotifydirectory watches fail over FUSE/sshfsand spam logs. Uncheck Enable real-time monitoring on each library and rely on scheduled scans instead. - Stale SSHFS mounts stalling scans at 0%: If a single
sshfssubmount drops into a staled??????????state, Jellyfin’sLibraryManagerstalls at0%during validation. Clearing it withfusermount3 -uz ~/mnt/nas3and remounting withidmap=userunblocks scanning immediately. - Point libraries to media subfolders: Adding top-level
/media/nas1or/media/nas3forces Jellyfin to walk non-media folders (like system backups or app binaries—which Jellyfin can even misclassify as.bindisc images). Pointing libraries directly at subfolders (e.g.,/media/nas1/Movies,/media/nas3/Movies) keeps scans fast and clean. - Jellyfin GUI folder picker: When adding libraries in Jellyfin’s web UI, FUSE submounts under
/mediadidn’t always appear in the folder browser. Typing the full path directly into the Folder path box worked.
3. Checking Tailscale Direct vs. DERP Relay
For smooth streaming, you want direct WireGuard connections rather than DERP relays. Running tailscale status on nas4 showed:
– nas2: direct (<remote-wan-ip>:41641)
– nas3: direct (<lan-ip>:41641)
– nas1: relay (DERP fallback)
Even with UDP 41641 open in firewalld on nas4, nas1 couldn’t hole-punch directly because its exit NAT remaps its outbound UDP port. nas2 and nas3 stream at full direct speed, while nas1 works over DERP until port forwarding is adjusted on nas1‘s router.
4. Monitoring Rootless Jellyfin in Weave Scope & Homelab Portal
Finally, I added a Jellyfin card pointing to http://<nas4-tailscale-ip>:8096 to the homelab_portal (nginx:alpine) running on nas3 so it’s one click away from anywhere on the tailnet.
To make the rootless jellyfin container visible in our 4-node Weave Scope topology as well, we enabled the user-level Podman socket (systemctl --user enable --now podman.socket) and mounted /run/user/<uid>/podman/podman.sock:/var/run/docker.sock into nas4‘s weavescope-probe instead of the root system socket.
Key Takeaways
- Rootless Quadlets +
:ro,rslavecleanly bridge FUSE and containers: You get systemd-managed auto-updates without running a privileged media server. - Verify
755permissions on remote FUSE roots:711directories work over SSH for owners but block non-root container UIDs reading throughsshfs. - Check
tailscale statusfordirectvsrelay: Knowing which NAS peers hole-punch directly immediately explains any remote buffering.