The :1 image runs Node 18 (EOL), which gets no fix for the October 2026
Node.js security release. :2 migrates kuma.db in place on first start;
back up kuma_data per host before pulling (see plan/10-uptime-kuma.md).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Phases 2 and 3 of plan/23-monitor.md. /api/summary adds: server online/MOTD/
version, players online with NMSR heads, registered-account count, disk free on
the world volume, backup freshness, join/leave history, and TPS.
Each tile has its own TTL cache and its own failure mode. A failed collector
keeps serving its last good value and reports the reason in a sibling *_err
field, so the page goes stale-with-a-warning rather than blank. The five
collectors refresh concurrently: sequentially, one request during an outage
would cost the sum of every timeout (~15s), and an outage is exactly when
someone is looking at this page. Measured with a dead dependency: 0.24s.
TPS needs RCON, so the monitor now holds RCON_PASSWORD. Read-only survives
because the allowlist is a compile-time constant and no HTTP input reaches a
command string — the UI cannot run commands because the server cannot be asked
to. The Source RCON client is implemented in-repo rather than added as a
dependency precisely to keep that property local and auditable.
Two things the real server corrected:
- `neoforge tps` prints "Overall: 20.000 TPS (0.357 ms/tick)", not the older
Forge "Mean tick time / Mean TPS" wording the plan assumed. main_test.go
pins the real captured output so a format change fails a test instead of
silently zeroing the tile.
- Rosters come from mc-status, which already holds the Drasl admin credentials
and exposes a sanitized list, so this container does not get a second copy.
Also fixes a phase-1 bug: the UI's level regex assumed itzg's console timestamp
([12:34:56]) but the file uses [01Aug2026 22:55:34.358], so every line was
classified INFO and the ERROR/WARN chips did nothing. It now matches on shape.
The container runs as root deliberately — /data is drwxr-x--- owned by uid 1000,
so a non-root user cannot traverse it and the tailer would silently read
nothing. The :ro mounts and absent port are what bound the damage.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Phase 1 of plan/23-monitor.md: a small Go service that tails the Minecraft
server's latest.log into an in-memory ring buffer and serves it to an embedded
single-page UI (autoscroll, substring/regex filter, level chips).
Read-only by construction — /data is mounted :ro, there is no mutating route,
and the container holds no credentials. It publishes NO port: caddy's
basic_auth vhost is the only way in, which is the point, since the service has
no auth of its own and the log carries player IPs, chat and admin commands.
Two non-obvious choices, both to avoid prod-only failures:
- Polling, not SSE. The chain is host nginx -> caddy -> app, and nginx buffers
proxied responses by default, so an EventSource would stall on cochi (whose
vhost is an untracked per-host file) and not reproduce locally. Delay was
accepted at design time, so polling is strictly better.
- The cursor is a sequence number, not a byte offset, and the tailer reopens on
inode change or size shrink. itzg rotates latest.log on every server restart,
i.e. every deploy — a naive tailer goes silent afterwards while still
answering HTTP 200. Verified against rotate, truncate, delete/recreate,
partial-line and ring-overflow cases.
MONITOR_ENABLED (on|off) gates both layers per host, mirroring FILES_BROWSE.
MONITOR_PASSWORD_HASH must be bcrypt AND single-quoted in .env: Compose
interpolates $-sequences in .env values, so a bare hash is silently mangled
($2a$14$abc... -> $$2a$$14...) and every login fails with no hint why.
Confirmed empirically, and it is the opposite of filestash, which needs $6$.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Per-host toggle for the read-only, no-auth browse of the shared upload dir,
so it can stay on for the LAN-party host (oxps) and be off on the
internet-facing one (cochi). Defaults to `on` — existing hosts unchanged.
Neither layer has conditionals, so each gets the value substituted in:
- caddy: docker-compose mounts 62-files.${FILES_BROWSE:-on}.caddy at the
fixed conf.d target (62-files.caddy -> 62-files.on.caddy, plus a
comments-only 62-files.off.caddy stub — caddy needs a parseable file
there). Same trick as FILES_MOUNT_MODE.
- nginx: render-nginx.sh splices nginx/files-browse.server.tmpl into
${FILES_BROWSE_SERVER} and adds/omits the name in the port-80 redirect.
This is the gate that matters; caddy is bound to 127.0.0.1.
Applying a change needs render-nginx.sh --install plus a caddy
--force-recreate (single-file bind mount pins the inode). With `off`,
drop `files` from LE_SUBDOMAINS on that host. upload. is unaffected.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Drop :ro from ./server-mods:/mods and the four kubejs source subdirs, and
add a new ./server-config:/data/config bind — mod config used to live only
inside the mc_data volume, unreachable from the host without root.
The config bind needs seeding: a bind mount shadows what is under it, so an
empty ./server-config hides the generated config and every mod falls back to
defaults. tooling/export-server-config.sh copies /data/config out of the
volume first, halting if the target is non-empty, if the container is still
running, or if the mc_data volume name is ambiguous.
Writable mounts also need the server's uid (itzg runs as 1000) to have write
access on the host dir; oxps' host user is 1002, hence MC_UID/MC_GID.
The kubejs ROOT still must not be mounted — BaseProperties.save() NPEs and
kubejs never initialises. That is about mounting the root at all, not about
the mode, so dropping :ro does not change it. It does mean kubejs writes now
land in $DISTRIBUTION_WEB_ROOT as uncommitted changes in that host's checkout.
Also documents how to reach the mc_data volume (exec/cp, resolved host path,
stop-before-write rule) in plan/05-minecraft.md.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Filestash cannot be served under a subpath (its SPA hardcodes /api, /assets,
/login at root), so it moves to its own subdomain upload.${BASE_DOMAIN}. The
apex-level ask was a static, no-auth file browser at files.${BASE_DOMAIN}: caddy
now file_server browse's the same ${FILES_DATA_DIR} share read-only there.
- filestash general.host -> upload.; nginx upload. vhost keeps the 2g body +
upload streaming + /admin throttle; new files. vhost is a plain static proxy.
- caddy: 60-files.caddy split into 60-upload.caddy (filestash) + 62-files.caddy
(file_server browse of /srv/files); caddy mounts FILES_DATA_DIR :ro.
- LE_SUBDOMAINS += upload (both upload. and files. are vhosts needing certs).
Security: files. is unconditionally public read (no auth), independent of
FILES_AUTH. The htpasswd login only gates writes via upload. Documented.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add an env-driven `user:`/`group_add` to the filestash service plus
tooling/filestash-set-run-user.sh. Running the container as a real host
account (oxps: oier=1002) makes every file it writes into the share owned
by that account, so scp/ssh and the web UI share one owner — no chown dance.
Defaults FILES_UID/FILES_GID=1000, leaving cochi (image's own uid) untouched.
group_add 1000 keeps the 0730 image binary + 0770 /app/data usable under any
uid; the helper also chowns the filestash_state volume (1000:1000) so a
non-1000 uid doesn't crash-loop unable to write its own sqlite/config.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add generators.${BASE_DOMAIN}: a pinned fork of misode.github.io served
with its own mcmeta + vanilla-mcdoc mirror, so it works with zero internet
at the LAN party. Fork lives in ulicraft-generators (branch ulicraft),
pinned at upstream 2bfe5522.
The app is 100% client-side: every data fetch comes from the player's
browser, not from cochi. So offline capability is entirely about which
origins the bundle refers to, plus a mirror to serve them from. The vhost
sets Content-Security-Policy with connect-src 'self' as ENFORCEMENT: an
external origin that sneaks back in (or returns via an upstream merge)
then breaks loudly in prod, on an ordinary day, instead of silently at the
venue with no internet to debug it.
sync-generators-data.sh mirrors ~109M: the 1.21.1 summary/data/atlas tags,
assets-tiny (assets minus hundreds of MB of .ogg — hence no sounds
generator), vanilla-mcdoc symbols.json, and summary/versions/data.min.json,
which comes from a BRANCH head rather than a version tag and runs during
Spyglass init. It verifies a sentinel file per source instead of trusting a
clean clone exit, since a renamed upstream path would otherwise sync
"successfully" and only fail in the browser at the party.
The container serves the bundle and the mirror on one origin, so there is
no CORS, no second cert and no second DNS name.
DEPLOY ORDER MATTERS. GENERATORS_REPO must be set in cochi's .env BEFORE
pulling this: compose uses ${GENERATORS_REPO:?...}, and an unset required
variable makes docker compose fail to parse the file — taking down every
service including Minecraft, not just this one. Full runbook and the rest
of the prerequisites: plan/21-generators.md.
Verified offline in real Chromium with every non-origin request aborted:
zero off-origin attempts, zero same-origin failures, zero console errors,
generators rendering schemas and live JSON.
One shared login (htpasswd) for all players, writable. Config is rendered
from a template and mounted read-only, so git stays authoritative — the
admin console loads but cannot save, by design.
Filestash's OIDC/SAML middlewares are enterprise-only, so Keycloak SSO is
out; FILES_AUTH keeps htpasswd/passthrough switchable for later. Auth-off
(passthrough) is only valid once the host is off the public internet —
render-config.sh warns loudly when rendering it.
Three traps, all found by testing against the pinned image:
- the `local` backend is admin-gated: without LOCAL_BACKEND_SECRET in the
connection params every login fails with "backend error - Not Allowed"
- the htpasswd plugin only accepts $2a$ bcrypt, so a $2y$ hash from
`htpasswd -B` fails every login silently — use `openssl passwd -6`.
render-config.sh rejects the wrong format rather than shipping it
- APPLICATION_URL/ADMIN_PASSWORD env make filestash rewrite config.json at
boot, which fails on the :ro mount — both live in the rendered config
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
online-mode=false made the server skip the online-auth handshake entirely, so
authlib-injector never validated sessions against Drasl: players got offline-hash
UUIDs and no skin textures (everyone Steve). authlib-injector's whole purpose is
to redirect the *online* handshake from Mojang to Drasl, so online-mode must be
true; only the 1.21+ secure profile stays off (ENFORCE_SECURE_PROFILE=false +
Drasl SignPublicKeys=false). Correct the conflation in CLAUDE.md and
plan/05-minecraft.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mounting the whole kubejs dir read-only made KubeJS's BaseProperties.save()
throw NPE on boot (it writes config/cache/generated/logs under kubejs/),
so KubeJS never initialised. Bind only the source subdirs (server_scripts,
startup_scripts, data, assets) read-only and let the kubejs root live
writable in mc_data.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The minecraft container only received forgemods jars via ./server-mods.
The distribution's files/kubejs (server_scripts, data, startup_scripts)
and CustomSkinLoader never reached the server, so server-side kubejs
recipes/scripts were silently absent.
Bind-mount both from DISTRIBUTION_WEB_ROOT (read-only, to avoid the
server polluting the client distribution with kubejs-generated files).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- landing header + footer brand: swap the Creeper pixel-art for the
ulicraft-logo-mini.svg goat mark.
- ServerCard fallback icon: same logo (real SLP favicon still overrides on
a successful ping).
- minecraft: set ICON=/extras/server-icon.png + OVERRIDE_ICON so the
in-game server-list entry and the SLP favicon use the goat logo. PNG
ships in ./runtime (mounted /extras).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Every avatar rendered the default skin. NMSR fetched the player profile from
Drasl fine, but the profile embeds an absolute HTTPS skin URL (Drasl BaseURL is
https), and NMSR could not connect to auth.${BASE_DOMAIN}:443: the mcnet alias
resolves auth. -> caddy, which only listens on :80 internally. Logs showed
`fetch_texture_from_mojang … client error (Connect)` for the skin URL, so NMSR
fell back to the default skin.
Mirror what the minecraft service already does: pin auth.${BASE_DOMAIN} to the
host via extra_hosts so NMSR reaches the host nginx on :443 (real LE cert,
publicly trusted — no private CA needed), and switch the nmsr mojank endpoints
from http:// to https:// so the profile/texture scheme matches the embedded
skin URLs.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Execute plan/18 (WS1-6) + plan/17.
- WS1: homepage join = 3 steps off packwiz (register -> UlicraftLauncher
-> join); per-OS downloads from a synced launcher-manifest.json (schema +
example committed); Fjord moved to its own /fjord page. Drop packwizUrl.
- WS2/3: mc-status gains an isolated /players endpoint (admin login ->
Drasl GET /players, own client+TTL+token cache, never blocks the status
path); online + registered rosters render shared name+avatar tiles;
AVATAR_MODE env (default headiso).
- WS4: /account skin CRUD via browser-direct Drasl (login -> upload ->
reset), in-memory token, players[0]; register relinks "Manage it here".
- WS5: footer link to status.${BASE_DOMAIN} in every footer.
- WS6: tooling/build-modlist.py generates mods.json + extracted logos from
the distribution forgemods; ModList.astro renders a flat list; stat tile
fed from the count.
Manifest/mods loaders read at build with a process.cwd() fallback (the
import.meta.url-only path does not resolve in Astro's bundled build).
New env: AVATAR_MODE, DRASL_ADMIN_USERNAME, DRASL_ADMIN_PASSWORD (mc-status
only). Generated mods.json / launcher-manifest.json / logos are gitignored;
.example.json files document the shapes. Build: 12 pages; mc-status builds.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The distribution-fed modset is built against NeoForge 21.1.233
(distribution.json); several mods hard-require it (ferritecore ≥21.1.218,
farmersdelight ≥21.1.219, configured ≥21.1.211). The old 21.1.209 pin made
the server crash in pre-load with "Missing or unsupported mandatory
dependencies: neoforge".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Drop packwiz entirely. The server now loads a filtered mod subset from a
local ./server-mods volume instead of fetching a packwiz pack at boot.
New flow:
- tooling/classify-mods.py (tomllib only, no network) reads each jar's
META-INF/neoforge.mods.toml and seeds mods-sides.json with a side
(client|server|both; default both). Existing entries are preserved.
- mods-sides.json: committed, keyed by jar filename, hand-editable override
surface. Initial seed: 65 both, 11 client, 0 server (76 jars).
- tooling/sync-server-mods.sh mirrors the both/server jars from
$DISTRIBUTION_WEB_ROOT into ./server-mods/ (authoritative; prunes strays;
halts on a missing jar). Replaces render-pack.sh.
- compose: minecraft mounts ./server-mods:/mods:ro with REMOVE_OLD_MODS=TRUE
(itzg auto-syncs /mods -> /data/mods). Removed PACKWIZ_URL, the pack.
extra_host (auth. kept for authlib), and depends_on caddy (drasl kept).
Both classify-mods.py and sync-server-mods.sh resolve their source dir as:
CLI arg / MODS_DIST_DIR env / $DISTRIBUTION_WEB_ROOT (in that order).
Removed: tooling/render-pack.sh + add-custom-mod.sh (packwiz tooling),
pack/ (packwiz source), custom/ (76 jars now sourced from the distribution),
the pack. caddy vhost + its mounts/alias, and the packwiz .gitignore block.
Client distribution is UNCHANGED: HeliosLauncher still installs the full
modset from distribution.json. Docs (CLAUDE.md, plan/04/05/14, runtime)
updated; plan/04-packwiz.md stubbed as superseded by plan/04-mods.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add a featured live ServerCard to the landing (replaces the static
ServerListPanel in hero + status sections): server favicon, color MOTD,
online/offline pill, players online/max with fill bar, and a player-head
row rendered by our own NMSR from Drasl skins. Progressive enhancement —
SSG skeleton degrades gracefully when JS or the API is unavailable.
Back it with a self-hosted pinger instead of the public api.mcstatus.io:
mcstatus.io's API service is closed-source (only the mcutil library is
open), so docker/mc-status wraps mcutil and re-emits its v2 JSON shape,
keeping the frontend unchanged. The service ignores the path address and
only pings MC_STATUS_TARGET (no SSRF relay), with a 30s TTL cache.
Exposed same-origin via caddy at the apex /api/mcstatus/* path (no new DNS
subdomain or LE cert change, no CORS). Uptime Kuma stays the uptime
history + alerting backend; see plan/15-mc-status.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Containers don't inherit the host's /etc/hosts; the LAN resolver returns a
dead address (192.168.0.3:443) for the public names, crash-looping the
packwiz install on boot. extra_hosts host-gateway routes HTTPS through the
host's nginx (valid LE cert) -> caddy -> service.
Also add plan/10-uptime-kuma.md (Minecraft monitor: internal + public probes)
and document the HTTPS resolution variant in plan/00-overview.md.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The ./custom:/srv/pack/custom mount nests inside the read-only ./pack:/srv/pack
mount. Docker can't mkdir the child mountpoint under a :ro parent, so on a fresh
checkout (no pack/custom/ dir) container init fails with "read-only file system".
Track an empty pack/custom/ so the mountpoint always exists.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Collapse the override-file matrix into a single docker-compose.yml holding the
whole stack: drasl, minecraft, mc-backup, caddy, nmsr, uptime-kuma. Bring-up is
now just `docker compose up -d --build`.
Removed features (dead-ends for this deployment):
- airgap / full asset mirror (mirror-airgap.sh, mirror-mods.sh, 20-mirror.caddy,
caddy local-CA export, /ca.crt)
- dnsmasq + avahi/mDNS + dns-records.sh + mdns-host-setup.sh + ENABLE_MDNS;
DNS is now handled outside this repo (HOST_LAN_IP dropped)
- Blessing Skin auth variant (compose + 00-core-blessingskin.caddy + BS_* env)
- host-nginx-static variant (docker-compose.nginx.yml, nginx/ulicraft.conf.tmpl)
- build-stack.sh and its run modes (online/airgap/core)
Production ingress is the only path now: host nginx terminates TLS (LE certs)
in front of caddy on a localhost-only port; caddy routes every vhost by Host
header. Launcher downloads move mirror/launcher -> launcher/.
Docs: README, deploy skill, and plan/ rewritten to match; obsolete plan docs
(dnsmasq, blessing-skin, assets-mirror, full-airgap-mirror, dns-and-run-modes)
deleted.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Caddy is no longer in the base stack. Two ingress paths now:
- docker-compose.caddy.yml — LAN/air-gap; build-stack.sh adds it for all
up modes + the prep pre-bake. Holds the mcnet auth./pack. aliases.
- docker-compose.nginx.yml — production behind the host's nginx: publishes
drasl on 127.0.0.1:25585, points minecraft at the host (extra_hosts) over
HTTPS for pack + authlib.
Base loses the caddy service and minecraft's caddy depends_on (re-added by
caddy.yml). Blessing Skin override now requires caddy.yml (it only swaps
caddy's core conf).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Move pack/custom/ -> ./custom/ and bind-mount it into caddy at
/srv/pack/custom (served URL pack.${BASE_DOMAIN}/custom/ unchanged).
Keeping jars outside the packwiz pack root stops `packwiz refresh`
from indexing them as direct files on top of the mods/*.pw.toml
external refs (drops 9 bogus custom/*.jar entries from index.toml).
add-custom-mod.sh now writes jars to ./custom/. Trim the blessingskin
caddy override to just the conf swap (volumes merge by target).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Rename the pack-serving vhost/alias from packwiz.${BASE_DOMAIN} to
pack.${BASE_DOMAIN} (caddy conf, caddy network alias, PACKWIZ_URL).
Inline the static-site caddy mounts (10-static.caddy, www, launcher,
CA cert) into the base compose; set MOTD to "Uli LAN party".
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
DNS is now your own party server. tooling/dns-records.sh prints the records
(online = service names; airgap = + Mojang/launcher/NeoForge spoofs) in several
formats. build-stack --up online|airgap selects the mirror layer and prints the
matching records. dnsmasq moved to an optional override (docker-compose.dnsmasq.yml).
Default domain switched to .lan (.local is hijacked by mDNS on unicast DNS).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The spoof aliases (maven.neoforged.net, libraries.minecraft.net, etc.) hijacked
those hostnames for every mcnet container, so the ONLINE NeoForge install during
pre-bake hit the empty mirror (cert path failure). Moved them to the mirror
override; core/static modes resolve the real upstreams.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Base compose is core only (dnsmasq, drasl, minecraft, mc-backup, caddy
auth+packwiz). docker-compose.static.yml and docker-compose.mirror.yml layer
on the landing site and the air-gap mirror independently. build-stack.sh --up
[core|static|mirror|full] selects the override files.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Mounting ./mirror/launcher at /srv/www/launcher failed: /srv/www is a
read-only mount, so the nested mountpoint can't be created. Mount it at
/srv/launcher and route /launcher/* via handle_path instead.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
oierandClaude Opus 5.5