OS Name/Version:
Ubuntu Server 24.04.2 LTS
Product Name/Version:
AMP Release “Phobos” v2.6.1.6
Problem Description:
I’m using AMP to host a Palworld dedicated server. When I issue the save
command via the AMP console (or /save
in-game), it responds with “Complete Save.” However, if I manually stop the server shortly afterward, I consistently lose 4–6 player levels and hours of world-building progress.
From reviewing logs and save file timestamps, it appears AMP is not ensuring the game has actually flushed its save data to disk before stopping the application. Since Palworld runs on Unreal Engine 5 — which performs saves asynchronously — the shutdown likely interrupts that process, resulting in data loss.
This issue is repeatable and not tied to any crash or improper shutdown. AMP should implement a post-save delay, or verify file write completion (e.g., via timestamp or log) before proceeding with shutdown.
Reproduction Steps:
- Play Palworld for several hours
- Use
/save
in-game orsave
in AMP console - Wait approximately 30–60 seconds
- Click “Stop” in AMP to shut down the server
- Restart the server
- Progress is rolled back significantly
Actions Taken to Resolve So Far:
- Verified
save
(without slash) is the correct AMP console command - Issued both
/save
in-game andsave
via AMP - Waited 30–60 seconds before clicking “Stop”
- Monitored file timestamps in
Saved/SaveGames/
— they often don’t update after a save - Reviewed AMP logs to confirm save was acknowledged before shutdown
- Created manual backup after saving — backup still contained outdated data
- Attempted to submit this report via the AMP UI (unsuccessful)
- Now reporting via CubeCoders community as the primary support channel
(Tags like palworld
, data-loss
, and amp
were not available when posting — please feel free to add them if appropriate.)
Update on Save/Backup Progress Loss Issue – May 19
After further testing and reviewing AMP’s event log in detail, I can confirm this is not a configuration issue — it’s a critical failure in how AMP or Palworld (or both) handle saves before backup/shutdown.
What Happened:
- On May 19, I was at level 25 in Palworld after hours of gameplay.
- I issued the
save
command in-game via/save
and again in the AMP console usingsave
. - I waited 2–3 minutes before taking a backup (
5-19-25
) while the server was still running. - After confirming all players had logged out, I stopped the server via AMP.
- I restarted the server — and my level was rolled back to 22.
- I attempted to restore the backup (
5-19-25
) — still level 22. - I even manually replaced the
.sav
files — only to get pushed back to level 21. - Eventually, after restoring the same AMP backup again, I landed back at 22. But my level 25 progress was permanently lost.
From the AMP Event Log:
AMP clearly shows that all relevant .sav
files (Level.sav, LevelMeta.sav, Player saves) were uploaded or manipulated after the backup was triggered. Here’s a snippet:
08:04 AM - Issued `save` via AMP console
08:05 AM - Server stopped via AMP
08:13 AM - Backup 5-19-25 created
08:34–08:35 AM - Save files manually uploaded via SFTP
08:36 AM - Restored backup 5-19-25
Despite this, the restored save file was still stale, missing multiple hours of playtime and several levels. This suggests AMP’s backup system either:
- Captures save data before it’s flushed to disk, or…
- Backs up from a cached location, or…
- The
save
command is misleading and doesn’t write to disk reliably before backup is triggered.
Final Notes:
- My server runs in a VM on ESXi using NFS with
sync=always
to a TrueNAS box — no write caching or async write risks. - This isn’t a system fault — it’s a save flow logic issue.
- It’s a huge deal because it makes Palworld unreliable in AMP, especially for multiplayer servers with real progress at stake.
Please:
- Add an AMP-level fix to confirm disk write completion before backup or shutdown.
- Consider forcing a configurable delay post-save.
- Or verify timestamp changes on
Level.sav
before proceeding.
I can provide full logs, SFTP timestamps, and screenshots if needed. This is a server killer and a trust-breaker — please prioritize a fix or recommend a stable workaround for now.