ARK servers crash. AsaApiLoader plugins crash even more. AASM v0.8.3 watches every server you've added, detects when the process is gone (not just suspended, not just slow to respond), and starts it again - usually within 30 seconds of the crash.

v0.8.3 also fixed a class of false-positive "CRASH CONFIRMED" loops that affected AsaApiLoader fleets under load (BUG-13 - the validator now classifies each check into gone / wrong-exe / timeout / parse-fail instead of treating all four as a crash).

Prerequisites:

A server already added to AASM.

Step 1 - Open Maintenance & Management

Server tab → Maintenance sub-tab → Maintenance & Management. The Quick Settings card on the left holds the crash-detection toggle.

Step 2 - Flip "Auto Restart on Crash" on

In the Quick Settings card, find the toggle labelled 🔧 Auto Restart on Crash. Click it on. AASM saves the setting immediately - no separate Save button.

This setting is per-server. Enable it on every server you want auto-recovered. The dashboard's server card shows a small "Crash Detection: ON" badge when it's active.

Step 3 - How AASM decides a server has crashed

v0.8.3 uses a multi-signal verification routine before declaring a crash:

Process check via WMI: is the ArkAscendedServer.exe (or AsaApiLoader equivalent) still running, with the right working directory?
If yes - server is fine. If no - go to step 3.
Consecutive-failure counter: AASM needs three failed checks in a row before triggering recovery. One missed check is just network jitter or RCON momentary backoff.
Bulk-empty guard: if the WMI query suddenly returns zero processes for ALL servers in your fleet (which usually means PowerShell/WMI itself hiccuped, not that everything crashed at once), AASM skips the cycle and re-checks on the next interval.

Step 4 - Crash-detection pauses (planned maintenance)

AASM auto-pauses crash detection during these operations so it doesn't fight the user-initiated action:

Manual stop / restart from the dashboard
SteamCMD updates
Settings Snapshot restores
Move-server / clone-server operations

The pause record is stored in data/crash_detection_pauses.json. v0.8.3 also restores crash detection automatically after the operation completes (BUG-17 fix from v0.8.2 - earlier builds would sometimes leave it paused on disk).

Step 5 - Tuning (optional)

For most servers the defaults are right. If you have an unusually slow PC or heavy mod load and you see false-positive crashes despite the v0.8.3 guards, open data/maintenance_settings.json for that server and increase the check interval. Don't change the consecutive-failure threshold - lowering it brings back the false-positive class.

If crash detection seems noisy on an AsaApiLoader server specifically, check the AsaApi log for plugin-init failures - the plugin can fail in a way that leaves the parent process running but useless. v0.8.3 surfaces this through the dashboard's server status, not the crash banner.