v1.13.0 — per-user access, and adoption that brings the saves

Ten changes since v1.12.1. Thanks to @Calliban001, @kajusnau, @excessnet, @jdvieira and @cproxy for the reports that drove most of them.

Operators and viewers can be limited to their own servers

An admin can now grant a user specific servers and clusters instead of the whole panel. A restricted user sees only what they were granted, plus every member of a granted cluster, and stops receiving other servers' logs and RCON output over the websocket — not hidden in the UI, never sent. A server they can't see returns a not-found page rather than an error that confirms it exists.

Everyone keeps full visibility until you restrict someone, so nothing changes on upgrade by itself. Two things do change:

• A new account with no role picked is an operator now, not an admin.
• The last admin can no longer be demoted or deleted. You'd otherwise lock yourself out of your own panel.

Adopting an existing server no longer calls it Euro Truck Simulator 2

Adoption worked out a container's game from its image repository and ignored the tag. Three games ship off ich777's SteamCMD image, so every container from that wrapper matched the last one declared, and a Palworld or V Rising server came back as a truck simulator with truck simulator settings. The lookup matches the tag too now, and wrapper tags Palisade has no support for stay out of the adopt list instead of being guessed at.

Adoption brings the saves across from ich777 Palworld and V Rising servers

The other half of the same report. Even correctly identified, adoption copied nothing between different images: it paired the two containers by matching mount paths, and ich777 keeps everything under /serverdata while Palisade's images use /palworld and /mnt/vrising. Nothing lined up, so the new server came up empty and the only way across was moving files by hand.

Palisade now knows where those two images keep their saves and copies them into the layout its own image expects, owned by the user that image runs as. Game files stay behind — both images reinstall them via SteamCMD on first boot. For V Rising, a save under any name other than world1 is renamed so the server actually loads it.

Adoption still only works for the image Palisade runs for a game, plus those two. Anything else stays out of the list, and the panel says why.

The connect address is the game server's, not the panel's

The address on a server's connect card came from whatever you had typed to reach Palisade. That is only the game server's address when the two share a network stack, which an Unraid install breaks easily: a manager on a custom or macvlan network has its own LAN IP, while a host-networked game server answers on the host's. Reported by @kajusnau, whose card advertised an address nothing was listening on while the one that worked was never shown.

A container can't see the host's LAN IP, so Palisade now works the address out from what you already configured: the new Address players connect to setting under Settings → General, then your port-forward target IP, then the host of your public base URL. If none of those is set it falls back to the browser address as before, so nothing changes for an install that configured none of it.

If you have port forwarding set up, you already told Palisade the LAN IP your game servers bind on, and every connect card now uses it. No action needed.

The appdata folder is called palisade now

The default App data path for new installs is /mnt/cache/appdata/palisade. It was /mnt/cache/appdata/ark-manager, left from when the project had that name, which meant you had to know the project's history to find your own files. Suggested by @kajusnau.

Your install is not affected and nothing moves. Existing containers keep their path, Unraid keeps your copy of the template, and updating changes nothing. Staying on ark-manager costs nothing.

The one case to know about: if you delete the container and reinstall fresh from the Apps tab, you'll get the new default and Palisade will start empty against a new folder while your data sits in the old one. It looks like data loss and isn't — set App data back to your existing folder. The README's "Moving off ark-manager" section has the steps if you'd rather move the directory instead.

UniFi joins pfSense for port forwarding

Settings → Integrations → Port forwarding now asks which router you have. UniFi Network needs a console host, API key, site and LAN target, with a Test connection button before you save. Per-server forwards work the same either way, and installs with pfSense already configured carry on untouched.

One change that applies to both routers: deleting a server now also removes the WAN forwards Palisade created for it, unless another server still needs the port. The delete dialog tells you before you confirm.

Backups reach the off-box copy, and Icarus can load a prospect

Off-box replication failed with tar exited 2 when retention deleted the oldest snapshot while the archive was still being written, and the backup then sat unreplicated until the hourly reconcile picked it up. Retention now waits its turn, so a new backup reaches the remote immediately. Failure messages carry tar's own error instead of just the exit code.

Icarus has a Load prospect on start setting under Session. The image blanks LoadProspect= on every boot, so hand-editing ServerSettings.ini never stuck; the value belongs in the setting now, and Palisade reapplies it. Names take letters, digits, spaces, dots, dashes and underscores.

Also

• scripts/deploy-unraid.sh changes the image and nothing else. It used to hard-code the network, ports, mounts and Docker access it recreated your container with, which moved installs onto the old ark-net name and switched them onto a socket proxy unasked. It now reads all of that back off the container it's replacing, and takes a container name if you run more than one.
• No more repeating could not inspect network ark-net ... 404 in the debug log on installs that were never on that network.
• Dependency bumps to clear the audit gate: Next 15.5.25, sharp 0.35.4, multer 2.3.0.
• CI and publish workflows moved onto current action majors before GitHub drops Node 20.

Upgrading

Update the container. One database migration runs at boot; it adds columns and tables for per-user access and changes no existing data. The appdata default changed for new installs only — see above if you plan to reinstall from Community Applications rather than update.