Mirrors Factorio's world feature for the SE server. The SE docker image always
loads the flat config/World dir and its entrypoint rewrites <LoadWorld> on every
start, so repointing the cfg is useless. Instead inactive worlds live in a
sibling config/Worlds library (not mounted) and switching is an instant mv-swap;
the active world name is tracked in .gsm-worldname and files are chowned to
1000:1000. New worlds are created from the image's built-in CustomWorlds
templates or by cloning the active world. Switch/clone require a stopped server.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Three compounding causes made the dashboard and detail pages slow:
- Detail page fetched the WHOLE fleet (every SSH/Prometheus/RCON call for
all servers) just to show one. Switch ServerDetail to GET /servers/:id
(add api.getServer) so it costs one server's worth of work, not N.
- No server-side caching: every 10s poll, from every open tab, recomputed
everything. Add a shared 5s-TTL per-server cache + in-flight dedup behind
both GET / and GET /:id; invalidate on start/stop/restart so power
actions still reflect immediately.
- Prometheus fetch had no timeout, so a slow/restarting Prometheus hung the
whole aggregation indefinitely (the .catch can't catch a hang). Add
AbortSignal.timeout to queryPrometheus/queryPrometheusRange.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The 15m access token only refreshed on a 401, but the dashboard uses optionalAuth and returns 200 even with an expired token, so the live page never noticed expiry until a manual reload.
Backend: ACCESS_TOKEN_TTL 15m -> 1h. Frontend: export refreshAccessToken; App.jsx schedules a refresh ~1min before the token expires (and on tab focus/return), and logs out cleanly (redirect to /login) if the refresh token is gone — no reload needed.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
getServerStatus returned 'offline' for both a stopped server AND an unreachable host. The dashboard hides 'unreachable' but shows 'offline', so an off VM oscillated in/out of the list as the failed-host breaker expired and re-tripped. Now distinguish: connection failure/hang -> 'unreachable' (host down), reachable-but-process-stopped -> 'offline'. Both autoshutdown and the Discord bot already treat non-online as not running.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Cached SSH connections had no keepalive, so when a VM/host is powered off the socket stays 'connected' (~2h OS TCP timeout) and execCommand hangs with no timeout, stalling status aggregation and leaving the dashboard on a stale 'online'.
Add keepaliveInterval/keepaliveCountMax to ssh.connect (dead peers detected in ~15s). Wrap getServerStatus/getProcessUptime in an 8s timeout that drops the dead cached connection and trips the failed-host breaker, so the fast 'unreachable' path kicks in. Affects all servers (V Rising, Palworld, etc.).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
config.json, discordBot.js, serverUpdates.js and ServerCard.jsx updated from Biohazard to FTB StoneBlock 4. Docs (infrastructure/gsm/CLAUDE) updated and docs/minecraft-modpack-install.md added. No Discord update notification is triggered (minecraft supportsUpdate=false).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
A validate pass alone does not unstick SteamCMD when it lands in state 0x6
with BytesToDownload=0 and a TargetBuildID newer than buildid: Steam never
schedules a depot fetch. Verified manually that deleting appmanifest_298740.acf
forces a clean reconciliation and a full re-validate that actually downloads
the missing depot. The handler now retries with the manifest removed when the
first pass leaves a pending target.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Steam was reaching state 0x6 with BytesToDownload=0, a known stuck-mid-update
pattern that resolves when SteamCMD is told to validate. Also stop relying on
the "Success!" stdout marker, which is missing in many real successes — compare
the buildid in appmanifest before and after instead.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The mmmaxwwwell SE image calls steamcmd with +login before +force_install_dir,
which Steam rejects ("Please use force_install_dir before logon!"), so container
restarts never actually update the game. The new handler runs SteamCMD in a
sibling one-shot container with the correct argument order, mounting the same
volumes as the live container, then leaves the server stopped for the operator
to start. The version cache is invalidated after a successful update so the UI
shows the new build immediately.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Reads the SteamCMD appmanifest_298740.acf via SSH and exposes
buildid + LastUpdated through a new /api/servers/:id/version endpoint.
The overview tab now renders a "Server Version" card and surfaces a
pending update hint when TargetBuildID differs from the installed build.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Bisher: 7d JWT im localStorage. Jeder 403 loggte den User aus — auch "Insufficient
permissions"-Responses fuer Background-Requests — wodurch User sich effektiv taeglich
neu einloggen mussten.
Backend:
- Neue refresh_tokens Tabelle (SHA256-gehasht, mit Rotation + Revoke)
- /auth/refresh tauscht Refresh-Token gegen frischen Access-Token (rotiert beide)
- /auth/logout revoked den Refresh-Token
- Access-Token-TTL auf 15m; Guest-Refresh 7d, Discord-Refresh 90d
- Daily cleanup-Job fuer abgelaufene/revoked Tokens
Frontend:
- api.js refreshed bei 401 automatisch (single-flight) und retryt die Anfrage
- 403 wird nicht mehr als Session-Ablauf behandelt
- Discord-Callback + Guest-Login liefern beide Tokens
- App.jsx synced React-State per gsm_token_refreshed CustomEvent
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Use real SE dedicated log patterns:
- Join: 'OnConnectedClient <Name> attempt'
- Leave: 'User left <Name>' (successful) / 'User left [<steamid>]' (failed attempt)
Unresolved leaves pop the oldest pending entry so failed connection attempts
don't stick in the player list.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Integrates Space Engineers dedicated server (docker+wine on 192.168.2.78) into GSM:
- config.json entry with docker runtime and log-based player detection
- getSpaceEngineersPlayers() parses docker logs for connect/disconnect events
- Extends type-check sites in rcon, autoshutdown, discordBot, servers routes
- Frontend ServerCard/ServerDetail logo + address wiring
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Switched Minecraft server from All the Mods 10 to Biohazard: Project Genesis
(Beta 0.4.5, MC 1.20.1 + Forge 47.4.0). Updated server name, Discord bot
display, frontend modpack link, and documentation.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add findSteamCmd() helper that searches common locations
- Provide path hints per server type as fallback
- Remove hardcoded paths, more robust for different setups
- Check /usr/games/steamcmd, ~/steamcmd/, ~/Steam/ etc.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Create generic SteamCMD handler factory for reusable update logic
- Check build IDs via appmanifest files and Steam API
- Support V Rising (1829350), Palworld (2394010), Zomboid (380870), tModLoader (1281930)
- Show update button for all supported server types in UI
- Minecraft, OpenTTD, Hytale marked as manual-update-only
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Container status 'created' means it was never started yet (e.g. after
docker create during update), so it should show as offline.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replaced emoji icons with actual server logos from gsm.zeasy.dev
as embed thumbnails for a more professional look.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Changed detection pattern from "hat das Spiel verlassen" to "ist weg"
to match actual Terraria server log format.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add getTerrariaPlayers function in ssh.js for PM2 log parsing
- Support German and English join/leave messages
- Update rcon.js to use Terraria log parsing
- Add Terraria to player fetch conditions in servers.js
- Update autoshutdown.js and discordBot.js for Terraria support
- Update config path to tModLoader directory
- Add global error handlers in server.js
- Update CLAUDE.md with deployment rules and Terraria info
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add hytale to serverDisplay config with icon 🏰 and color
- Fix fetchServerStatuses to get players for Hytale (no RCON)
- Enables player join/leave alerts for Hytale
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Add server.type === 'hytale' check to player count API routes,
matching the fix in autoshutdown.js.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Hytale uses log parsing instead of RCON, so add explicit check for
server.type === 'hytale' to enable player count checking.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Join: [World|*] Player 'Name' joined world ... (UUID)
- Leave: [PlayerSystems] Removing player 'Name (Name)' from world ... (UUID)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add getHytalePlayers function to ssh.js that parses server logs
- Track player joins via [Universe|P] Adding player pattern
- Track disconnects via Removing player pattern
- Integrate with rcon.js getPlayers and getPlayerList functions
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add missing "hytale" entry to SERVER_JOBS in prometheus.js
- Fix config editor regex bug that was corrupting CSS class names (e.g. text-cyan-400 becoming text-cyan-<span>400</span>)
- Use placeholder tokens to prevent numbers regex from matching inside already-replaced spans
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add Node Exporter target for Hytale server (10.0.30.204:9100)
- Add Hytale config read/write functions to ssh.js
- Add GET/PUT /hytale/config API routes
- Create HytaleConfigEditor.jsx with JSON syntax highlighting
- Add Hytale config tab to ServerDetail.jsx
- Add stopCmd and port to Hytale server config
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add tmux runtime support to ssh.js (status, start, stop, logs, uptime)
- Add Hytale server configuration to config.json
- Add Hytale server info and logo to frontend (ServerCard, ServerDetail)
- Add Hytale emoji to Discord notification mapping
- Update documentation with Hytale server details
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Adds POST /api/servers/discord/send-update endpoint that allows
superadmins to send announcements to all Discord guild update channels.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Database files should not be in version control as they contain
production data that gets overwritten on each deployment.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add isGuest flag to UserContext
- Block guests from navigating to /server/:id route
- Make ServerCards non-clickable for guests
- Add rejectGuest middleware to backend
- Protect server detail endpoints (/:id, /metrics/history, /whitelist)
Guests can now only view the dashboard overview without accessing
individual server details.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove temp files and reorganize docs
- Add .gitea/workflows/deploy.yml for automated deployment
- Add unreachable host checks to server routes (/:id, logs, start/stop/restart)
- Add unreachable checks to config routes (zomboid, terraria, openttd)
- Return HTTP 503 with unreachable flag instead of crashing
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
