Hytale Server Lost Progress After a Crash? Saves and Rollbacks

Published on

Why a Hytale server can revert to its last boot after a crash, which log lines and save settings to check, and how --backup, /backup and /stop protect a world.

Three Hytale characters next to a wooden village notice board covered in pinned notes

Your Hytale server went down, and when everyone joined again the last session was gone: all the progress reverted to when the server booted up last time. Here is why a server can lose a whole session, what to check, and how Hytale's own backup options keep the next crash small. It all comes from Hypixel's server manual, API reference and patch notes and the current release build, server 0.7.5 of 29 October 2026, checked on 3 November 2026.

Quick answer: A Hytale server writes changed chunks and player data to disk all the time while it runs, so a crash on its own should only cost the last moments before it. If your progress reverted to when the server booted, the server could not save during that session. Search the log for Failed to save chunk, check the free disk space, and check that IsSavingChunks and IsSavingPlayers are true in the world's config.json. Then turn on the built-in backups with --backup --backup-dir backups (every 30 minutes by default), run /backup before anything risky, and stop the server with /stop so it finishes saving before it exits.

How a Hytale server saves while it runs

Saving does not wait for shutdown. Hypixel's API reference lists a chunk save queue that is checked every 0.5 seconds (QUEUE_UPDATE_INTERVAL) and a world save interval of 10 seconds (World.SAVE_INTERVAL), and changed chunks are written in the background. A clean stop adds a final pass: each world logs Saving Chunks... and then Done Saving Chunks before the process exits.

That gives you two very different kinds of loss:

What you lostWhat it points to
The last moments before the crashThe process died before its last writes. Expected when a server crashes or is killed
Everything since the last bootSaving failed, or was switched off, for the whole session

Why a server reverts to its last boot

Saving failed all session. When a chunk cannot be written, the server logs Failed to save chunk (x, z): with the reason on the next lines. While the server runs you notice nothing. When the process dies, the world comes back from disk, and the disk only holds what was saved before the errors started. A full disk does this, and so does missing write permission on the universe folder.

Saving is switched off for the world. Each world has a config.json in universe/worlds/<world>/, and the manual's example shows "IsSavingChunks": true and "IsSavingPlayers": true. Set to false by a plugin or an admin, they stop that world saving chunks or player data. With chunk saving off, the shutdown log says This world has opted to disable chunk saving so it will not be saved on shutdown.

A /worldgen reload left saving locked. The Update 8 pre-release notes of 29 October list this fix: "Fixed /worldgen reload leaving a world unable to save until a restart when the chunk delete step ended early." It is not in a release with notes yet, so restart a release server after you use that command.

A world crashed and was reloaded. If a world's thread crashes, players on it see "The world you were on has crashed". On a dedicated server the default world is then reloaded "from disk" (WorldCrashRecovery), with up to 3 attempts 5 seconds apart, and if that fails the server shuts down (CrashRecoveryConfig). A reload shows what the disk holds, so this is often when a broken save comes to light.

The process was killed instead of stopped. kill -9, a force kill button, an out-of-memory kill or a power cut ends the server without its final save. With working background saves that costs seconds. Without them, it costs everything since the last successful write.

Check the log first

Each start opens a new log file in the logs folder of the server's working directory, named after the start time, such as logs/2026-11-03_18-02-11_server.log. The restart made a fresh file, so the session that crashed is in the file before the newest one. On Linux, from the folder the server runs in (Server/ with the official scripts):

cd /opt/hytale/Server
ls -t logs | head -3
grep -n -E "Failed to save chunk|opted to disable chunk saving|crashed|Scheduled backup|Backup failed|Done Saving Chunks" logs/2026-11-03_18-02-11_server.log

On Windows, Select-String -Path "logs\2026-11-03_18-02-11_server.log" -Pattern "Failed to save chunk" in PowerShell does the same for one search term.

Log lineWhat it tells you
Failed to save chunk (x, z):A write failed. The next lines name the cause
This world has opted to disable chunk saving...IsSavingChunks is false for that world
Reloading crashed world '<name>' in 5 second(s) (attempt 1 of 3)A world thread crashed and is being reloaded
Scheduled backup to run every 30 minute(s)Automatic backups are on, at that interval
Backup failed with exceptionA backup failed. The next lines say why
Done Saving ChunksThat world finished its final save on a clean stop
No shutdown lines, the log just endsThe process was killed or the machine went down

Check the disk and the save settings

On Linux, df -h . shows the free space and du -sh universe backups logs shows what uses it. RocksDB worlds also keep a write-ahead log of up to 512 MB on release builds, which the Update 8 pre-release cuts "from 512 MB to 128 MB".

Then confirm IsSavingChunks and IsSavingPlayers are true for each world. On the running server, /world settings chunksaving shows the current value and /world settings chunksaving set true turns it back on, and playersaving works the same way. From the console, add --world <name>. If you edit config.json instead, stop the server first, because the manual warns that "Manual changes while the server is running are likely to be overwritten."

On our Hytale hosting the disk runs from 20 GB of NVMe storage on Small to 50 GB on Extreme, and with full file access you can download old backup zips and clear space yourself.

Turn on Hytale's built-in backups

The server has its own backup system. You switch it on with start flags or with a Backup block in the server's config.json, and the flags win when both are set (BackupConfig):

Start flagconfig.json key in BackupDefaultWhat it does
--backupEnabledoffTurns on scheduled backups
--backup-dir <path>Directorynot setFolder for the zips. Without it, nothing runs
--backup-frequency <minutes>FrequencyMinutes30Minutes between backups, at least 1
--backup-max-count <n>MaxCount5Recent backups to keep
--backup-archive-max-count <n>ArchiveMaxCount5Archived backups to keep

The official start.sh and start.bat already pass --backup --backup-dir backups --backup-frequency 30, so with those scripts your backups land in Server/backups. If you start the server another way, look for Scheduled backup to run every 30 minute(s) near the top of the log. No line means no backups. To start it yourself with a backup every 15 minutes, from the Server folder:

java -XX:AOTCache=HytaleServer.aot -jar HytaleServer.jar --assets ../Assets.zip --backup --backup-dir backups --backup-frequency 15

Or stop the server and add this block at the top level of config.json:

"Backup": {
  "Enabled": true,
  "Directory": "backups",
  "FrequencyMinutes": 30,
  "MaxCount": 5,
  "ArchiveMaxCount": 5
},

With the official scripts, their --backup-frequency 30 beats FrequencyMinutes. To change the interval there, edit the DEFAULT_ARGS line, and check it after updates, because the updater replaces the scripts.

What the server does with these settings:

  • The first backup comes one interval after boot. Nothing is backed up when the server starts or on a normal stop, so a crash 20 minutes after a restart leaves you with the backups from before it.
  • Each backup is a zip of the universe folder, named after the server's local time, such as 2026-11-03_18-30-00.zip. config.json, permissions.json, bans.json, whitelist.json and mods/ are not in it, so copy those yourself. Our admin guide explains what they hold.
  • The copy is consistent. Hypixel's server code waits for pending world saves and holds new ones while the zip is written.
  • Old backups rotate. Before each new backup, the server deletes all but the newest five. The oldest of those it removes goes to backups/archive instead if the newest archived copy is more than 12 hours older than it, and the archive keeps five.
  • The zips are not compressed. Each one is as large as your universe folder, and the defaults can hold up to eleven. Hypixel's hardware post put 5,000 by 5,000 explored blocks at about 661 MB, before Chapter I made saves smaller. Eleven copies of that are about 7 GB, so on a small disk lower MaxCount or move old zips off the server.

Keep --backup-dir out of the Hytale launcher's install folder, or the server exits with code 11 (see our Chapter I server guide). And download a copy now and then, because the backups sit on the same disk as the world.

Back up or save on demand

/backup makes the same zip right away. It needs a backup folder, from --backup-dir or from Directory in config.json. Without one, the server replies that it must be started with --backup-dir.

/world save --all writes every world to disk now, and /world save --world <name> saves one. Use it before a planned restart or a risky change.

Do not copy the universe folder of a running server with a file manager or FTP: the server keeps writing while you copy. Use /backup, or stop the server first.

Stop the server cleanly

  1. Type /stop in the console. Ctrl+C in the terminal that runs the server starts the same shutdown, because the server registers a shutdown hook with Java.
  2. Wait for Done Saving Chunks for each world and for the process to exit.
  3. Only then restart, update or copy files.

If your panel has both a stop and a kill option, use stop, and kill only a server that hangs. Big worlds need that time: the Update 8 pre-release fixes "a server shutdown problem where a slow world save could be cut short or run a second time".

Restore a world from a backup

Stop the server, pick the newest zip from before the problem (the names use the server's local time), and move the current universe folder aside instead of deleting it. The zip starts at worlds/ and players/, so unzip it into a new universe folder:

cd /opt/hytale/Server
mv universe universe-before-restore
mkdir universe
unzip backups/2026-11-03_18-30-00.zip -d universe
ls universe

On Windows, in PowerShell:

Rename-Item universe universe-before-restore
Expand-Archive -Path "backups\2026-11-03_18-30-00.zip" -DestinationPath universe

universe should now contain worlds and players, not another universe folder. Start the server, check your builds, and delete universe-before-restore once you are sure.

If a world will not load after a crash

Copy universe first, then start the server once with --verify-worlds, which Hypixel describes as "Verify all worlds and then exits":

java -jar HytaleServer.jar --assets ../Assets.zip --verify-worlds

If it finds damage, it logs Failed to verify world <name>, <broken>/<total> chunks corrupted and stops. Add --recovery-mode FROM_BACKUP_OR_REGENERATE (plus --backup-dir backups if your zips are there) to look for each broken chunk in your backup zips and regenerate it if no backup has it. --recovery-mode REGENERATE skips the backups, so those chunks generate fresh and lose what was built in them.

The crash policy itself lives in a CrashRecovery block, inside Defaults in config.json for the default world or in a world's own config.json, with the keys Mode (None, Reload or Shutdown), MaxAttempts, RetryDelaySeconds and Fallback. Other worlds have no recovery unless they carry that block.

FAQ

Why did my Hytale server reset to when it last started? Because it could not save during that session, for example on a full disk or with IsSavingChunks set to false. A crash alone only costs the last moments, since the server saves in the background.

Does a Hytale server save automatically? Yes. Changed chunks go to disk in the background, and a clean /stop saves once more. Backups are a separate feature that you turn on with --backup and --backup-dir.

How do I back up a Hytale server? Start it with --backup --backup-dir backups, as the official start scripts do, for a zip of universe every 30 minutes. Run /backup for one right away, and copy the config files and mods/ separately.

Where are Hytale server backups stored? In the --backup-dir folder, with older copies in its archive subfolder. With the official scripts that is Server/backups.

How do I restore a Hytale world from a backup? Stop the server, rename universe, create an empty universe folder and unzip the backup into it, so that it contains worlds and players. Then start the server.

Sources

All opened on 3 November 2026: