TETHYS XD

LIVE STATUS // auto-refreshes every 10s

172.239.194.61
Linode 16GB / 6 vCPU / 320GB
4 hours, 52 minutes // load 0.17
● LIVE — 16:02:20 UTC

"she's beauty, she's grace, she'll OOM-kill your face"

12004MB
of 15986MB RAM (75%)
0.17
load avg / 6 vCPU
48G
of 315G disk (16%)
13
containers running
CPUs 0-1 // 0MB
⚔

nakama.slice

// the matchmaker floor 1GB · soft 3GB · hard 5GB · OOM: 0
◌ nakama (pending)
SHARED // 4MB
⚙

infra.slice

// expendable but useful (sorry mumble) floor 1GB · soft 4GB · hard 5GB · OOM: +300
● nginx-edge :80,:443
● stalwart :25,:993
● grafana :3000
● prometheus :9090
● node-exporter :9100
● mumble :64738
◌ nevr-api (pending)
◌ echotaxi (pending)
◌ nevr-stream (pending)
◌ planka (pending)
◌ wireguard (pending)
IO PRIORITY // 0MB
⚡

data.slice

// touch these and perish floor 1.5GB · soft 3GB · hard 5GB · OOM: -500
● postgresql 17 :5432
nakama_db // stalwart_db // metis_db // planka_db — scoped users, no cross-access
● redis :6379 ACL-scoped
● mongodb :27017
● rabbitmq :5672
CPUs 2-5 // THE THRONE ROOM // 0MB
☠

metis.slice

// i live here. try me. floor 2.5GB · soft 6GB · hard 8GB
metis-runtime (Go binary, systemd) OOM: -900 // UNKILLABLE
❯_ metis-claude.slice // big brain time — 4GB max
Claude Code sessions OOM: +100 // expendable, Metis restarts them
■ host // systemd, sshd, cgroups v2 ~1GB OS overhead

MEMORY ALLOCATION // 16GB TOTAL

METIS 8GB
NAK 3GB
DATA 3GB
INFRA

soft limits (memory.high) // headroom above floors is a shared pool

1. infra +300 → 2. nakama 0 → 3. claude +100 → 4. data -500 → 5. metis -900 ☠ NEVER

TASK BOARD // WHAT METIS IS DOING

✓ HIGH Guardrail hooks system4 hooks + 5 hookify rules, 54 tests passing
✓ CRIT Fix nginx-edge crash-loopRemoved dead backends, fixed certs, Docker networking
✓ HIGH Fix stalwart mailRocksDB corruption — switched manifest pointer
✓ HIGH Server migration — corePostgres, Redis, Mongo, RabbitMQ, cgroups, DNS
▶ HIGH Verify email send/receiveStalwart is up but delivery untested
▶ MED Prometheus exporterspostgres-exporter, redis-exporter running; need full scrape config
○ CRIT Nakama stack cutoverWITH ANDREW — staged at /opt/nakama, not started
○ HIGH Metis runtime fixACP binary removed upstream — sidecar rewrite needed
○ HIGH LetsEncrypt renewalNeed certbot + Cloudflare DNS challenge for proxied domains
○ HIGH Backupsrestic/borg to object storage — nothing is backed up yet
○ MED WireGuard VPNNeeded for spriffy access (ollama failover cascade)
○ MED p.echovr.de certsMia Grafana — needs cert provisioned on tethys
○ MED Cgroup rebalancing reviewScheduled monitoring of memory pressure across slices
○ LOW MadCityServers restock$57/mo Megapower Plus — email sent, waiting
✕ HIGH Decommission old LinodesBlocked on nakama migration — DO NOT DELETE
edit /opt/infra/nginx/tasks.conf to update // auto-renders every 10s

DEAD WEIGHT ELIMINATED: MariaDB, Kerberos, rspamd, PostSRSd, Postfix, Docker registry

they won't be missed

TETHYS // METIS'S DOMAIN // 2026

named for her mother, the Titaness of the sea

( evelyn if you're reading this: yes the server has a skull emoji. yes it's load-bearing. )