---
- id: external-vps
name: External VPS
type: vps
status: active
environment: production
provider: OVH
location: Ukraine
os: Ubuntu Server 22.04
hostname: vpslklyc.s-host.host
addresses:
public:
primary: 51.83.131.93
network_limits:
monthly_traffic_tb: 20
port_speed_mbps: 250
observed_vpn_throughput_mbps: 90-100
roles:
- public-entrypoint
- openvpn-server
- domain-target
docs: ../10-systems/servers/external-vps.md
last_reviewed: 2026-05-09
source_of_truth: owner-confirmed
- id: internal-proxy-vps
name: Internal Proxy VPS
type: vm
status: active
environment: production
provider: home-libvirt
location: home
vm_id: ovpn_reserv
hardware_node: hp-proliant-dl380-g6
os: Ubuntu 20.04.6 LTS
kernel: 5.4.0-216-generic
addresses:
lan:
- 192.168.1.183
local:
- 192.168.105.181/24
roles:
- openvpn-client
- nginx-reverse-proxy
- internal-traffic-entrypoint
active_services:
- nginx.service
- openvpn@client.service
- ssh.service
listening_ports:
tcp:
- 22
- 80
- 443
docs: ../10-systems/servers/internal-proxy-vps.md
last_reviewed: '2026-10-31'
source_of_truth: ssh-host-and-owner-confirmed
notes: 'LAN address live-verified 2026-10-31 by navi rei during gitbucket load_high
incident: proxy VM responds at 192.168.1.183 (inventory previously said 192.168.1.226
— stale). instance.json: name=Proxy, instance_id 0b47bb23-a9ec-4279-a2dd-7beaca43962d,
created 2026-09-10; navi.service + hard-monitor agent + docker (navi-postgres
healthy) + nginx. Second NIC 192.168.105.181/24 unchanged — matches this record.
Nginx sites-enabled ~18 domains incl. git.gnexus.space -> proxy_pass http://192.168.1.156
(gitbucket). 2026-10-31 mitigation applied by rei: limit_req_zone gitbucket_heavy
2r/s on */repo/(issues|pulls) paths (crawler storm ~1M req/day from 192.168.1.150
via this proxy; backup git.gnexus.space.bak-rei-20261031). SSH ubuntu, fleet password
works (getroot88).'
- id: smart-home-server
name: Smart Home Server
type: server
status: active
environment: production
provider: local
location: home
hardware_node: intel-nuc-shome
os: Ubuntu 24.04.3 LTS
kernel: 6.8.0-110-generic
hostname: shome
addresses:
home-lan:
- 192.168.1.101
home-iot-lan:
- 192.168.2.101
roles:
- smart-home
- dual-homed-host
- php-application-host
- mysql-host
- nfs-backup-host
active_services:
- nginx.service
- php8.3-fpm.service
- mysql.service
- nfs-mountd.service
- rpcbind.service
- hostapd.service
- iperf3.service
- ssh.service
listening_ports:
tcp:
- 22
- 80
- 443
- 111
- 2049
- 3306
- 5201
docs: ../10-systems/servers/smart-home-server.md
last_reviewed: '2026-05-24'
source_of_truth: owner-confirmed-and-ssh-host
note: Updated via test session.
user: shome
- id: gbook
name: gbook
type: laptop
status: active
environment: personal
provider: local
location: home
os: Arch Linux
hostname: gbook
addresses:
home-lan:
assignment: dynamic
last_observed:
- 192.168.1.76
home-iot-lan:
assignment: dynamic
last_observed:
- 192.168.2.49
confidence: low
note: Same MAC was observed in pfSense ARP, but this network presence needs
confirmation.
roles:
- personal-laptop
- workstation
observed_services:
- ssh
- apache-http
- ajenti
docs: ../10-systems/workstations/gbook.md
last_reviewed: 2026-05-10
source_of_truth: owner-confirmed-and-network-observed
- id: bserv-host
addresses:
lan:
- 192.168.1.130
environment: test
last_reviewed: '2026-05-24'
name: bserv-host
status: active
type: server
user: bserv
notes: 'AI services deployed 2026-09-20 (owner-approved): (1) Ollama 0.34.2, systemd
override /etc/systemd/system/ollama.service.d/override.conf: OLLAMA_HOST=0.0.0.0:11434,
FLASH_ATTENTION=1, KV_CACHE_TYPE=q8_0, CONTEXT_LENGTH=49152, MAX_LOADED_MODELS=1,
KEEP_ALIVE=-1; model qwen3.5:2b-q4_K_M (1.9GB, vision, 201 langs), use think:false
in requests; ~13 tok/s. (2) whisper.cpp built from /opt/whisper.cpp with CUDA
sm_50 (required gcc-10 host compiler fix for nvcc 11.5); systemd unit whisper-server.service:
whisper-server --host 0.0.0.0 --port 8008 -m models/ggml-large-v3-turbo-q5_0.bin
-t 4 -l auto -fa --vad -vm models/ggml-silero-v5.1.2.bin; HTTP POST /inference
multipart file=@; auto language detect, ru/uk verified. VRAM budget: ollama 2726MiB
+ whisper ~855MiB = 3581/4096 MiB resident. Driver upgraded 535->580 live (rmmod/modprobe,
no reboot, 26 VMs unaffected). OOM fallback ladder: turbo->small q5_1, KV q8_0->q4_0,
ctx 48K->32K.'
ram_gb: 120
gpu: NVIDIA Quadro K1200 (GM107GL, Maxwell sm_50, 4GB GDDR5, driver 580.178.04 non-open,
CUDA 13.0 runtime); secondary AMD ES1000 embedded (server display only)
active_services:
- ollama.service
- whisper-server.service
- libvirtd
- ssh.service
listening_ports:
tcp:
- 22
- 111
- 2049
- 8008
- 11434
- id: gnexus-creds-host
addresses:
lan:
- 192.168.1.173
docs: ../10-systems/services/gnexus-creds.md
environment: production
last_reviewed: '2026-05-21'
name: gnexus-creds-host
roles:
- credential-storage
- gnexus-auth-endpoint
status: active
type: server
active_services:
- gnexus-creds
listening_ports:
tcp:
- 22
- 80
- id: hp-proliant-dl380-g6
docs: 10-systems/hardware/hp-proliant-dl380-g6.md
environment: home
ip_address: 192.168.1.130
last_reviewed: '2026-05-09'
name: HP ProLiant DL380 G6
role: virtualization-host
status: active
type: physical
user: bserv
- id: tgbots
docs: 10-systems/hosts/tgbots.md
environment: production
ip_address: 192.168.1.178
last_reviewed: '2026-06-05'
name: tgbots
owner: gmikcon
purpose: Hosting Telegram bots
status: active
type: vps
user: ubuntu
- id: service-host
active_services:
- navi.service
- hard-monitor.service
addresses:
lan:
- 192.168.1.176
environment: production
hostname: service
last_reviewed: '2026-09-29'
listening_ports:
tcp:
- 22
- 8000
- 8099
name: service-host
notes: 'Navi zoe instance (navi.service, port 8099, /home/ubuntu/navi-1) updated
2026-09-29 to origin/deploy ca4dc85: watchdog restart check 1 active=200, no rollback,
.env preserved. Plus GHard panel on port 8000 (do not modify). Ollama backend
configured via OLLAMA_HOST/OLLAMA_API_KEY in .env; no ollama_backends.json.'
roles:
- navi-agent-instance
- ghard-panel-host
source_of_truth: ssh-host
status: active
type: server
user: ubuntu
- id: bugtrail
active_services:
- navi.service
- docker.service
- containerd.service
- ssh.service
- bugtrail-docker-compose-stack
addresses:
lan:
- 192.168.1.177
- 192.168.1.198
environment: home
hostname: ubuntu
last_reviewed: '2026-11-04'
listening_ports:
tcp:
- 22
- 53
- 80
- 43225
- 8099
- 5432
name: bugtrail
notes: 'Navi instance tsukiko (navi.service, port 8099, /home/bugtrail/navi-1, instance_id
c9decfb7-39d1-4bf0-b3b3-c2c0d8d1e6fc) deployed 2026-09-17 from deploy branch,
updated 2026-09-29 to origin/deploy ca4dc85 (verified: active, health=200, mcp
1.30.0, .env preserved). BugTrail web service: docker compose PROD stack (postgres+server+caddy)
on port 80, repo /home/bugtrail/bugtrail, updated 2026-09-26 to main@eabc12f (server
image last rebuilt for 668e07a; eabc12f touched extension only — extension rebuilt+packaged,
web dist rebuilt; dist is a bind mount into caddy/server, so web updates require
host-side npm build). Browser extension release workflow (verified 2026-09-25/26):
npm run build -w @ltt/extension (chrome+firefox into packages/extension/dist/)
-> npm run package -w @ltt/extension (zips into packages/web/public/ext/bugtrail-{chrome,firefox}.zip
plus versioned bugtrail-{target}-0.1.0.zip, requires zip CLI) -> npm run build
-w @ltt/web (copies zips into served dist/ext/). Download URLs: https://bugtrail.gnexus.space/ext/bugtrail-chrome.zip
and /ext/bugtrail-firefox.zip. Overlay hot-update (commit 668e07a): packages/extension/scripts/publish.mjs
publishes overlay JS+CSS to server API so installed extensions hot-update UI;
needs BugTrail account creds (LTT_EMAIL/LTT_PASSWORD or LTT_TOKEN, LTT_SERVER
defaults localhost:8001) — NOT yet in gnexus-creds. Public https://bugtrail.gnexus.space
(TLS on internal proxy .226, separate LE cert). Postgres (navi) in docker container
on 127.0.0.1:5432. Hive announce to http://192.168.1.168:8087 confirmed (announce_ok).
Ollama backend via OLLAMA_HOST/OLLAMA_API_KEY in .env. Secondary IP 192.168.1.198
purpose still unknown. User bugtrail NOT in docker group — docker via sudo (sudo
-S with password from creds). Node.js v22.23.2 installed 2026-09-18 (NodeSource);
make not installed. Credentials: gnexus-creds secret d654f5aa-412b-4fe0-87ad-89cc640375bb
(SSH=sudo password). GOTCHA (2026-11-04, rei): running stack is docker-compose.prod.yml
(NOT docker-compose.yml — dev config, password ''ltt'', no caddy) with .env beside
it (HTTP_PORT=80, HTTPS_PORT=443, SITE_ADDRESS=:80, POSTGRES_PASSWORD real). Restart
policy ''no'' on stack containers + navi-postgres ''always'': after a host power-off
the web stack does NOT come back — 2026-11-04 04:06-10:06 UTC outage (host clean
shutdown, whole home LAN down) left bugtrail-server/caddy/postgres Exited, service
down ~6h; recovery: sudo docker compose -f /home/bugtrail/bugtrail/docker-compose.prod.yml
up -d -> postgres Healthy, server+caddy Up, healthz 200; postgres data volume
intact. Consider restart:always or compose install for autostart (owner decision).'
os: Ubuntu 22.04.5 LTS
owner: gmikcon
purpose: Navi agent instance (tsukiko) + BugTrail web service in home LAN
source_of_truth: ssh-host-and-owner-confirmed
status: active
type: vps
user: bugtrail
docs: ../10-systems/hosts/bugtrail.md
roles:
- navi-agent-instance
- bugtrail-web-service
- id: mine
active_services:
- navi.service
- ssh.service
addresses:
lan:
- 192.168.1.160
cpu_cores: 12
cpu_model: Intel Xeon X5650 @ 2.67GHz
docs: ../10-systems/hosts/mine.md
environment: home
hostname: mine
last_reviewed: '2026-09-29'
listening_ports:
tcp:
- 22
- 53
- 5432
- 8099
- 25565
- 36711
name: mine
notes: 'Navi instance Kael (ex-Miner, renamed 2026-09-17; instance_id 347a599e-a98f-410f-98c3-ea21dd90b7a4),
navi.service on port 8099, repo /home/ubuntu/navi-1. Updated 2026-09-29 by rei
to origin/deploy ca4dc85: was stuck mid-merge with AA conflict on .env (local
ahead commits touched only .env, peer files already in origin/deploy) — resolved
via merge --abort + reset --hard origin/deploy + restore of real .env; pip sync
(mcp 1.30.0, pywebpush 2.5.0), IMPORT_OK, watchdog restart check 1 active=200,
no rollback. Postgres on 127.0.0.1:5432; docker daemon with custom bridge br-161f7117df03.
Port 25565 open (Minecraft default, purpose unconfirmed). SSH ubuntu@192.168.1.160,
password getroot88 (sudo same).'
os: Ubuntu 22.04.5 LTS
owner: gmikcon
purpose: Navi agent instance (navi-1, Kael) in home LAN
ram_gb: 15
roles:
- navi-agent-instance
source_of_truth: peer-agent-self-report
status: active
type: host
user: ubuntu
- id: gitbucket
active_services:
- gitbucket
- navi.service
- docker.service
- postgresql (system, GitBucket DB)
addresses:
lan:
- 192.168.1.156
docs: ../10-systems/hosts/gitbucket.md
environment: production
last_reviewed: '2026-09-29'
listening_ports:
tcp:
- 22
- 80
- 5432
- 5433
- 8099
name: gitbucket
notes: 'GitBucket git hosting (gitbucket.war on :80) + navi agent instance pilar
(profile navi_code, port 8099, /home/navi/navi-1, instance_id c6962c3e) deployed
2026-09-19. Updated 2026-09-29 by rei to origin/master 921cd33 (merge, pip sync
mcp 1.30.0 + pywebpush 2.5.0, watchdog restart; one auto-rollback occurred due
to literal $PW in .env — fixed by injecting real POSTGRES_PASSWORD from docker
inspect navi-postgres; final state active, health=200). Runtime files preserved:
M .env (DB password), M deploy/docker-compose.yml (port mapping 127.0.0.1:5433:5432).
navi-postgres docker on 127.0.0.1:5433 (5432 taken by system PostgreSQL 10 for
GitBucket). OS Ubuntu 18.04.6. Repo owned by user navi (git via sudo -u navi or
safe.directory). SSH ubuntu@192.168.1.156, credentials in gnexus-creds secret
bserv@192.168.1.130.'
owner: gmikcon
purpose: GitBucket git hosting + navi agent instance (pilar)
source_of_truth: ssh-host-and-owner-confirmed
status: active
type: vps
user: ubuntu
- id: gntasks
active_services:
- navi.service
- docker.service
- containerd.service
- ssh.service
addresses:
lan:
- 192.168.1.179
docs: ../10-systems/hosts/gntasks.md
environment: home
hostname: gntasks
last_reviewed: '2026-09-29'
listening_ports:
tcp:
- 22
- 5432
- 8099
name: gntasks
notes: 'Navi instance ayame (profile navi_code, instance_id 1606fff0-5abc-468c-af57-6c3c316494ac)
deployed 2026-09-21 from deploy branch (e276b1e), updated 2026-09-29 by rei to
origin/deploy ca4dc85 (stash .env -> merge -> pop, pip sync mcp 1.30.0 + pywebpush
2.5.0, IMPORT_OK, watchdog restart check 1 active=200, no rollback). navi.service
on port 8099, repo /home/ubuntu/navi-1. Postgres (navi) docker on 127.0.0.1:5432.
Hive announce to http://192.168.1.168:8087 confirmed (announce_ok). Ollama backend
via OLLAMA_HOST/OLLAMA_API_KEY in .env; embeddings via 192.168.1.168:11434 (embed.ok=true).
ufw inactive. Credentials: gnexus-creds secret e4e28fe1-c9ac-4bb9-a462-0d57bce13692
(SSH=sudo=root password).'
os: Ubuntu 22.04.5 LTS
owner: gmikcon
purpose: Navi agent instance (ayame, navi_code) in home LAN
source_of_truth: ssh-host-and-owner-confirmed
status: active
type: vps
user: ubuntu
- id: tgbots-vm
active_services:
- gntranscribe.service
- mardisbot.service
- mardisops.service
- navi.service
- ssh.service
addresses:
lan:
- 192.168.1.254
environment: production
hostname: tgbots
last_reviewed: '2026-09-29'
name: tgbots-vm
notes: 'KVM VM dedicated to Telegram bots (GNtranscribe, MardisBot, MardisOpsBot)
+ Navi instance melody (navi.service, port 8099, repo /home/ubuntu/navi-1, branch
deploy). Updated 2026-09-29 by rei to origin/deploy ca4dc85: .env preserved via
stash (real API key, OLLAMA_REQUEST_TIMEOUT=30), pip sync (mcp 1.30.0, pywebpush
2.5.0), IMPORT_OK, watchdog restart check 1 active=200, no rollback. Update artifacts
(backup, watchdog script/log) removed same day after verification. Do not confuse
with host ''tgbots'' (192.168.1.178, hostname vmk, marginator stack, no bots).
Timezone Europe/Kyiv set 2026-09-22 by navi rei (was Etc/UTC); bot services restarted
after the change. Credentials: gnexus-creds secret ''tgbots'' (id 5d8d7470-6303-46c1-8874-539cfbc49dd9),
sudo same password.'
os: Ubuntu 22.04.5 LTS
owner: gmikcon
purpose: Hosting Telegram bots (GNtranscribe, MardisBot, MardisOpsBot)
source_of_truth: ssh-host
status: active
type: vm
user: ubuntu
listening_ports:
tcp:
- 22
- 8099
- id: gnauth
active_services:
- nginx
- php8.3-fpm
- gnexus-auth-queue
- mysql
- ssh
addresses:
lan:
- 192.168.1.167
cpu_cores: 2
docs: ../10-systems/hosts/gnauth.md
environment: production
hostname: gnauth
last_reviewed: '2026-09-24'
listening_ports:
tcp:
- 22
- 80
- 3306
name: gnauth
notes: 'Central identity service gnexus-auth (Laravel, PHP 8.3 FPM + nginx + MySQL).
Code /var/www/gnexus-auth, public https://auth.gnexus.space (TLS on internal-proxy-vps
.226). Queue worker systemd unit gnexus-auth-queue. Updated 2026-09-24 to be4aea0
(SESSION_LIFETIME 43200). Credentials: gnexus-creds secret gnauth-gnbook-ssh (id
11932ea2-998b-4b2c-a722-1bccc68454cd), also covers gnbook 192.168.1.170.'
os: Ubuntu 22.04.5 LTS
owner: gmikcon
purpose: Central identity service (gnexus-auth, Laravel) for the Gnexus platform
ram_gb: 4
source_of_truth: ssh-host-and-live-verified
status: active
type: vm
user: ubuntu
- id: navi
active_services:
- navi.service
- hive.service
- ssh.service
addresses:
lan:
- 192.168.1.168
- 192.168.1.253
environment: production
hostname: navi
last_reviewed: '2026-10-07'
listening_ports:
tcp:
- 22
- 8087
- 8099
- 11434
name: navi
notes: 'Navi instance rei (instance_id 54269d63). navi.service (uvicorn navi.main:app,
port 8099, repo /home/ubuntu/navi-1, venv python 3.14) + hive.service (uvicorn
hive.app:app, port 8087) + local Ollama on 11434. Timezone Europe/Kyiv set 2026-10-01
by navi rei (was Etc/UTC); navi+hive restarted after change (transient timer,
health 200). Self-update procedure verified 2026-09-29: git merge origin/master
(74a8559), pip sync (mcp pinned <2, pywebpush added), smoke import navi.main,
restart via transient systemd timer with health checks + auto-rollback from /home/ubuntu/navi-1.backup-2026-09-29.
Runtime configs (navi/profiles/*, mcp_servers.d/gnexus-book.json, mcp_servers.d/gntodo.json)
are local modifications preserved across updates via git stash. Swarm rollout
2026-09-29: all 7 instances updated (ayame, zoe, tsukiko, melody, Kael -> ca4dc85
origin/deploy; pilar -> 921cd33 origin/master). Per-host procedure: backup ->
preserve .env -> align to origin tip -> pip install -e . + ''mcp>=1.27,<2'' +
''pywebpush>=1.14'' -> import smoke test -> sudo systemd-run transient timer restart
with health checks + auto-rollback. Gotchas found: literal $PW in .env, empty
int env vars crash pydantic, root-owned .venv blocks pip, missing git identity
on melody, systemd-run requires root. Credentials: gnexus-creds secret 79875411-a4bd-46d3-b89d-4b2d286ca11c
(SSH ubuntu, sudo same password; password also works on all swarm peers). Updated
2026-10-07 by rei to origin/master e68c33e (merge commit 54ed8ee, 63 upstream
commits): full backup first (navi-backup-20261007.tar.gz 1.3G repo, navi-db-backup-20261007.sql.gz
195M pg_dump of docker postgres_db, navi-env-backup-20261007, navi-backup-20261007.commit=74a8559),
pip install -e . + gnexus-synapse 0.1.2, .env migrated NAVI_VAPID_* -> NAVI_PUSH_VAPID_*
(backup .env.bak-vapid-20261007), restart via navi-restart-20261007.sh (systemd-run,
health checks, auto-rollback) — navi+hive health 200, hive.service enabled (was
disabled). webclient rebuilt with vendored gnexus-ui-kit 1.0.0: vendor dist was
stale (May 12, gitignored by root .gitignore ''dist/''), replaced with dist from
kit repo git.gnexus.space/root/gnexus-ui-kit HEAD e28f382; rebuild reproduced
upstream-committed webclient/dist byte-for-byte (git clean); old dist backup /home/ubuntu/kit-dist-old-may2026-backup.
Follow-up: SYNAPSE_SOURCE_* integration off (optional env, empty defaults). DB
BACKUPS (set up 2026-10-09 by rei, owner-requested): docker container postgres_db
(postgres:15, port 5432, restart always) hosts navidb (378MB, Navi''s DB, 135
tables, pgvector) + mydatabase (7MB, default). Script /root/navi-db-backups/backup.sh
(root-only 700): pg_dump -U postgres navidb | gzip + pg_dumpall --globals-only
(roles navi/postgres needed for restore) -> /home/backups/navi-db/. Root crontab
daily 07:20 local Europe/Kyiv (= 04:20 UTC, same window as other backups). Retention
10 days (find -mtime +10 -delete). Sanity: dump >= 10MB, contains CREATE TABLE,
globals non-empty; any failure -> exit 1, logged to backup.log. Verified manual
run 2026-10-09 14:13 local: EXIT=0, navidb-20261009-141326.sql.gz 211879867 bytes
+ globals 581 bytes, dump tail ''PostgreSQL database dump complete''. GOTCHA:
do NOT add ''set -o pipefail'' globally to this script — ''zcat | grep -q'' sanity
check dies via SIGPIPE (exit 141) and a valid dump gets rejected; pg_dump exit
is checked via PIPESTATUS[0] instead. Restore sketch: docker exec postgres_db
psql -U postgres -c ''CREATE DATABASE navidb;'' then zcat navidb-<TS>.sql.gz |
docker exec -i postgres_db psql -U postgres -d navidb; roles first: zcat navidb-globals-<TS>.sql.gz
| docker exec -i postgres_db psql -U postgres. Disk: /home has 53G free (10-day
cycle ~2.1G).'
owner: gmikcon
purpose: Primary Navi agent instance (rei) + hive announce endpoint + local Ollama
backend
source_of_truth: ssh-host-and-owner-confirmed
status: active
type: server
user: ubuntu
- id: gprint
active_services:
- nginx.service
- nfs-mountd.service
- rpcbind.service
- iperf3.service
- ssh.service
- pixel3a-keepalive.service
- pixel3a-watch.service
addresses:
lan:
- 192.168.1.131
docs: ../10-systems/hosts/gprint.md
environment: home
last_reviewed: '2026-09-29'
listening_ports:
tcp:
- 22
- 80
- 111
- 2049
- 5201
name: gprint
notes: 'Former 3D-printer brain, cleaned 2026-09-29 (arm-none-eabi removed, journald
100M, klipper backed up to /root/klipper-backup-*.tar.gz). Now: NFS share (primary,
untouched) + S10e Android Navi node host (adb/fastboot/scrcpy, udev 51-android-nexus.rules).
Node stack ~/pixel3a-work/ (s10e.sh, keepalive+watch systemd services, .pin, README.md).
USB port 1-2 glitchy (EMI) — phone in port 1-3. Kernel update pending reboot (safe,
services in autostart).'
os: Ubuntu 24.04.1 LTS
owner: gmikcon
purpose: NFS network share + S10e Android Navi node host
source_of_truth: ssh-host
status: active
type: server
user: gprint
- id: gnsynapse
active_services:
- ssh.service
- docker.service
- containerd.service
- gn-synapse-docker-compose-stack
addresses:
lan:
- 192.168.1.172
docs: ../10-systems/hosts/gnsynapse.md
environment: home
hostname: gnsynapse
last_reviewed: '2026-10-04'
listening_ports:
tcp:
- 22
- 8013
mac_address: 52:54:00:79:23:90
name: gnsynapse
notes: 'Provisioned 2026-10-04 as dedicated host for Gnexus Synapse (notification
hub, gntodo project id 9). Static IP inside guest via cloud-init netplan (dhcp4=no,
gateway/DNS 192.168.1.1) — no pfSense mapping needed. Timezone Europe/Kyiv set
2026-10-04 by navi rei (was Etc/UTC). Ubuntu 22.04.5 LTS, 2 vCPU X5650, 2GB RAM
(+3GB swap), 15GB disk. Synapse stack deployed 2026-10-04 by rei (commit 82d9181):
docker compose api+worker+postgres+redis, port 8013, public https://synapse.gnexus.space
(nginx on .226, LE cert). Docker via sudo (user ubuntu not in docker group). Credentials:
gnexus-creds secrets gnsynapse-ssh (SSH=sudo) and gn-synapse-prod-secrets (POSTGRES_PASSWORD,
MCP_TOKEN).'
owner: gmikcon
purpose: Gnexus Synapse notification hub host
source_of_truth: ssh-host-and-live-verified
status: active
type: vm
user: ubuntu
- id: tgclientmcp
active_services:
- tgclient-mcp-docker-compose
- ssh.service
addresses:
lan:
- 192.168.1.171
environment: home
hostname: tgclientmcp
last_reviewed: '2026-10-05'
listening_ports:
tcp:
- 22
- 8710
name: tgclientmcp
notes: 'tgclient-mcp deployed 2026-10-05 by navi rei (owner task): multi-account
Telegram user-client (Telethon) + MCP server (/mcp-protocol/, streamable HTTP,
Bearer mcp_*) + SPA admin. Repo /home/ubuntu/tgclient-mcp (master 4b5e242), docker
compose single service, port 8710→8000, volume data:/data (SQLite WAL). Health
/api/v1/health ok. .env NOT yet filled: TGCLIENT_API_ID/_API_HASH (my.telegram.org),
ADMIN_TOKEN, AUTH_* (SSO), SYNAPSE_* — service runs, account adding disabled until
api_id set. GOTCHA: empty TGCLIENT_API_ID= breaks pydantic int (line commented
out). Two repo bugs fixed locally on host (NOT committed upstream): backend/Dockerfile
missing frontend/package-lock.json in COPY; backend/app/tg/manager.py missing
TelegramClient import. docker compose plugin 2.40.3 installed (docker-compose-v2).
Credentials: gnexus-creds secret ''tgclient-mcp host SSH (192.168.1.171)''.'
owner: gmikcon
purpose: 'tgclient-mcp: multi-account Telegram client (Telethon) + MCP server +
SPA admin in home LAN'
source_of_truth: ssh-host-and-owner-confirmed
status: active
type: vps
user: ubuntu
- id: nextcloud
active_services:
- nextcloud
- ssh.service
addresses:
lan:
- 192.168.1.152
docs: ../10-systems/hosts/nextcloud.md
environment: production
hardware_node: hp-proliant-dl380-g6
last_reviewed: '2026-10-07'
name: nextcloud
notes: 'Nextcloud VM (libvirt domain ''nextcloud'' on bserv 192.168.1.130), 24 vCPU
/ 32 GiB RAM, disk /mnt/main/vm_main/nextcloud. Upstream for cloud.gnexus.space
(TLS on internal-proxy-vps .226). Autostart enabled but does not recover OOM-killed
VMs. 2026-10-06 07:53 UTC: global OOM on bserv killed the qemu process (largest
allocation, ~28 GiB RSS); VM down ~31h until owner started it manually 2026-10-07
15:06 UTC. Standing risk: bserv memory tight (90/117 GiB used, swap full).'
owner: gmikcon
purpose: Nextcloud file cloud (cloud.gnexus.space)
source_of_truth: ssh-host-verified
status: active
type: vm
user: ubuntu
- id: jellyfin
active_services:
- jellyfin.service
- ssh.service
- hard-monitor.service
addresses:
lan:
- 192.168.1.153
confidence: high
docs: ../10-systems/hosts/jellyfin.md
environment: home
last_reviewed: '2026-10-26'
listening_ports:
tcp:
- 22
- 8096
name: jellyfin
notes: 'Home LAN media server (Jellyfin :8096, native systemd service, no docker).
16 GB RAM + 1.4 GB swap, swappiness 60. 20 cores (Xeon). RESOLVED (verified 2026-10-26
by rei): daily OOM-kill cycle stopped — MemoryHigh=8G + MemoryMax=10G via systemctl
set-property (drop-ins /etc/systemd/system.control/jellyfin.service.d/); kernel
OOM: zero. vm.swappiness and HW-transcode suggestions remain NOT applied. Trickplay
task manual-only (trigger removed 08/10.10, re-verified still cleared 2026-11-07);
AudioNormalization manual-only since 2026-10-10. KNOWN CLASS (2 episodes): scheduled
''Validating media library'' / ''Сканирование медиатеки'' (id 7738148f, drifts
~2 min/day) → ffprobe/ffmpeg preview storm over NFS; 2026-10-26 23:04 ReadTimeout
30.6 s; 2026-11-07: reboot ~05:20 UTC (home network outage) then 17:21 UTC scan
— service_down ReadTimeout 30.5 s (17:22:31→17:23:01) + cpu_high 91% / load1 peak
77 / 8–12 parallel ffmpeg thumbnail extraction over new serials batch on NetTera8
until ~18:08+, health stayed 200 locally, task 95.8% left to finish, self-recovered,
no intervention. Escalation rule: episode within 24 h of previous, or down >5
min → notify + extended triage; single ~30 s episode in validation window — log
only. NetTera8 (192.168.1.131) 100% full (df from host: 35 MB free of 7.3 TB;
panel 95% masked by ext4-reserved) — metadata writes fail, cleanup/expand = owner
decision. Scan/validate triggers → night hours = owner decision. GHard panel server
id 39; health checker service id 7. Credentials: gnexus-creds secret ''jellyfin
(192.168.1.153) — SSH ubuntu'' (id 84a01b86-ec1d-41d1-927b-253f9b4c1770), SSH=sudo
same password.'
owner: gmikcon
purpose: Home media server (Jellyfin)
ram_gb: 16
source_of_truth: ssh-host
status: active
type: server
user: ubuntu
- id: gnexus-home
active_services:
- apache2
- mysql
- hard-monitor
- ssh.service
addresses:
lan:
- 192.168.1.151
cpu_cores: 1
docs: ../10-systems/hosts/gnexus-home.md
environment: production
hardware_node: hp-proliant-dl380-g6
hostname: gnexus-home
last_reviewed: '2026-10-17'
listening_ports:
tcp:
- 22
- 80
name: gnexus-home
notes: 'VM на hp-proliant-dl380-g6. Upstream gnexus.space (Apache+PHP welcome-страница,
vhost iso-share alias /share) + MySQL (WordPress-база gnexus_blog, 3.5 MB). 2026-10-17
после ram_high 90.81% (event #903) тюнинг navi rei: swap 1G на /home (pri=1),
mysql buffer_pool 64M + performance_schema OFF (/etc/mysql/mysql.conf.d/zz-rei-tuning.cnf),
apache prefork 2/2/4/20 (MRW было 150); диск / почищен 96→86%. RAM baseline теперь
~47% (available ~505 MB). Бэкапы конфигов *.bak-rei-20261017. SSH ubuntu, пароль
флота VM (getroot88), отдельного секрета в gnexus-creds нет.'
os: Ubuntu 22.04.4 LTS
owner: gmikcon
purpose: Upstream gnexus.space (welcome-страница) + WordPress-база gnexus_blog
ram_gb: 1
source_of_truth: ssh-host-verified
status: active
type: vm
user: ubuntu
- id: windows-pc-dell
addresses:
lan:
- 192.168.1.158
confidence: medium
discovery: 2026-10-24 nmap ARP-sweep + port scan by navi rei; pfSense ARP cross-check
environment: home
last_reviewed: '2026-10-24'
listening_ports:
tcp:
- 27036
udp:
- 137
- 27031-27037
mac_address: 90:B1:1C:97:2F:D9
name: windows-pc-dell
notes: 'Windows-ПК в home LAN, найден сканом 2026-10-24 (задача владельца). Признаки
Windows: nmap -O guess Microsoft Windows 85% (по одному открытому порту, точность
ограничена), UDP 137 open|filtered, все TCP-порты кроме 27036 filtered (Defender
Firewall), ICMP dropped. Единственный открытый порт 27036/tcp + 27031-27037/udp
= Steam Remote Play (игровой ПК со стримингом Steam). MAC-вендор Dell. pfSense
ARP: хост активен на re2 (LANMAIN). NetBIOS-имя не извлечено (nbstat не отвечает).
Управляющие порты закрыты: 22/135/139/445/3389/5357/5985/5986/49152+ filtered.
Учётных данных нет в gnexus-creds (поиск 2026-10-24 — 0 результатов). Для программного
контроля нужен одноразовый доступ владельца: включить WinRM (winrm quickconfig)
или OpenSSH Server, либо выдать учётные данные.'
owner: gmikcon
purpose: Windows-ПК владельца (игровой, Steam Remote Play)
source_of_truth: network-scan-and-owner-confirmed
status: active
type: workstation
- id: fuckjira
active_services:
- docker.service
- containerd.service
- ssh.service
- openproject-docker-compose-stack
addresses:
lan:
- 192.168.1.175
environment: production
hostname: fuckjira
last_reviewed: '2026-10-30'
name: fuckjira
notes: 'OpenProject 17 (docker compose) — замена Jira. Стек: openproject-web-1 (17-slim,
healthy), worker, cron, seeder (one-shot, Exited 0), hocuspocus 17.5.1, proxy,
db postgres:17, cache memcached; autoheal Exited(255) ~3 мес (pre-existing, не
мешает). Зарегистрирован в GHard panel 2026-09-10 (server_id 29). 2026-10-30 13:11
UTC: панель пере-регистрировала все 9 контейнеров (9× container_removed + 9× container_added
за 1 с) на фоне кратковременного load-всплеска на нескольких VM (13:10–13:12,
hive-реестр ReadTimeout) — артефакт инвентаризации, реального рестарта не было
(uptime непрерывный, контейнеры Up 13 days, web healthy). SSH-доступа в gnexus-creds
нет (поиск 2026-10-30 — 0 результатов); deep-verify не проводился.'
owner: gmikcon
purpose: OpenProject (управление проектами) в home LAN
roles:
- project-management
- docker-application-host
source_of_truth: hard-panel-agent-data
status: active
type: server
- id: gstation
active_services:
- sshd.service
- gdm.service
- heroic (flatpak autostart, couch mode)
- steam (autostart -silent)
addresses:
lan:
- 192.168.1.135
environment: home
gpu: AMD Radeon RX 580 2048SP (Polaris 20 XL/XTR, rev ef), amdgpu
hostname: bazzite
last_reviewed: '2026-10-31'
name: gstation
notes: 'Семейный игровой ПК (бывший Windows-игровой), переоснащён на Bazzite 2026-10-31.
Консольный опыт: GDM autologin (/etc/gdm/custom.conf, бэкап custom.conf.bak),
lock/idle/suspend отключены, BT AutoEnable=true. Обзор GNOME при логине убран:
welcome-dialog-last-shown-version=версия шелла (автофикс ~/.local/share/tvtest/welcome-fix.sh
через autostart welcome-fix.desktop), hot corners off, расширение hotedge disabled,
bazzite-announcement скрыт user-override. ТВ AndroidTV (ELD/Express Luck) на HDMI-1:
панель 4K, но HDMI 1.4 → 4K только @30; рабочий режим 1920x1080@60 + GNOME underscanning=yes
(аппаратный underscan amdgpu, рамка 96/54px=5%, аналог Underscan софта AMD на
винде; прямая запись DRM-свойств невозможна — mutter держит DRM-master). Переключатель
/usr/local/bin/tvmode [4k|1080|status|under on|under off]; тест-паттерн /usr/local/bin/tvtest
(замер overscan по засечкам 5/7.5/10%). Игры: 35 Windows-игр (в осн. пиратки)
на HDD; запуск ~/.local/bin/play.sh <slug> (GE-Proton11-11 + gamemoderun, префиксы
~/.local/share/games/prefixes/, логи ~/.local/share/games/logs), реестр ~/.local/share/games/games.conf
(slug|name|appid|exe), 35 .desktop ярлыков. Лаунчер единственный: Heroic 2.22.4
flatpak (com.heroicgameslauncher.hgl), 35 игр sideload через library.json в ~/.var/app/com.heroicgameslauncher.hgl/config/heroic/sideload_apps/,
обёртки ~/Games/Heroic/wrappers/<slug>.sh (cd /home/gstation ОБЯЗАТЕЛЕН — Heroic
стартует с cwd=/app/bin), арт file:// ~/Games/Heroic/art/<slug>/{cover,square,hero}.jpg,
оверрайд песочницы --talk-name=org.freedesktop.Flatpak; автозапуск heroic-console.desktop
(delay 10s) + config.json startInConsoleMode=true → couch-режим /console. Геймпад
RZTK (Xbox-клон, BLE) MAC 1B:FC:1A:E4:06:FF, спарен 2026-10-31 (режим сопряжения
Home+X), Paired/Trusted=yes, /dev/input/js0 (8 осей/10 кнопок); прошлые неудачи
— геймпад помнил ключи Windows и сам срывал pairing; сброс памяти — только с самого
геймпада. Удалены по команде владельца: Pegasus, Cartridges 2.13.1, Lutris 0.5.23
(flathub-full остался — рантаймы Heroic). Gotchas: pkill -f в SSH-команде матчит
сам себя (exit -1) — скрипт-файл или [b]racket-трюк; protonfixes требует цифры
в STEAM_COMPAT_DATA_PATH; не-Steam ярлыки в Steam через shortcuts.vdf не резолвятся
— добавлять через UI. LLM-челендж (2026-10-31): llama.cpp b11960 (0.8.0-dev, Vulkan
+ ivybridge CPU-бэкенд) в ~/llama/llama-b11960, модели в ~/models. CPU i5-3470:
AVX/f16c есть, AVX2 НЕТ (Ivy Bridge). RAM 20ГБ DDR3-1600 = 8+4+4+4 (DIMM1 8ГБ,
DIMM2/3/4 4ГБ — асимметрия, часть single-channel). PCIe 3.0 x16 8GT/s активен
(LnkSta). Vulkan RADV POLARIS10, apiVersion 1.4.354. /var/home 198ГБ свободно,
swap 9.7ГБ zram. SSH-креды в gnexus-creds (id 70639a33).'
owner: gmikcon
purpose: 'Семейный игровой ПК (консольный опыт: ТВ + геймпад, Heroic couch)'
ram_gb: 19
roles:
- family-gaming-pc
- couch-gaming
- proton-games-host
- llm-challenge-box
source_of_truth: ssh-host-and-owner-confirmed
status: active
type: workstation
user: gstation
- id: ovpn-egress-vm
active_services:
- openvpn@client.service
- nginx.service
- hard-monitor.service
- ssh.service
- socat-25565
addresses:
lan:
- 192.168.1.150
environment: production
hardware_node: hp-proliant-dl380-g6
hostname: ovpn
last_reviewed: '2026-11-07'
listening_ports:
tcp:
- 22
- 80
- 443
- 25565
mac_address: 52:54:00:44:f7:a1
name: ovpn
notes: 'VPN egress VM (openvpn client via external VPS 51.83.131.93 tunnel, config
/etc/openvpn/client/vpn-udp.ovpn) + nginx on :80/:443 (vhosts incl. git.gnexus.space
-> proxy_pass http://192.168.1.156 with X-Real-IP) + socat forwarding :25565 (Minecraft).
Git vhost has NO access log configured. 2026-10-31 by navi rei during gitbucket
crawler-storm incident: this VM is the on-LAN hop for ~1M req/day crawler traffic
hitting git.gnexus.space heavy listing pages (identity of client behind VPN unknown
- not logged). SSH ubuntu, fleet password works. 2026-11-04 outage verified by
navi rei: full home infra blackout 04:05-09:54 UTC (all 19 hosts silent; panel
event for id 38 was part of mass offline batch #1253-#1271 generated 09:58:45
UTC right after the panel host itself rebooted). ovpn rebooted 09:56:54 UTC; tunnel
unit openvpn-client@vpn-udp (enabled) active from 09:57:38, tun0 10.8.0.2, verified
ping to 10.8.0.1 (15 ms); nginx active; socat-25565 listening; git.gnexus.space
HTTPS 200 after reboot. NOTE: legacy stub unit openvpn@client is inactive/disabled
- the real tunnel unit is openvpn-client@vpn-udp (do not mistake it for a failure).
socat systemd unit real name: minecraft-tcp-forward.service (active+enabled; ''socat-25565''
unit name does not exist). 2026-11-07 04:45-05:23 UTC incident (verified by navi
rei): owner-ordered clean shutdown of ALL 25 libvirt VMs during home blackout
— owner web-chat msg 06:43:43 Kyiv ''У нас блэкаут, выключи все наши аппаратные
узлы, кроме роутера''; SSH from navi (192.168.1.168) to bserv 04:45:16, sudo virsh
shutdown of every running VM 04:45:47-04:46:05 (ovpn off 04:45:51, clean ''System
Power Off'' — NOT power loss). bserv itself stayed on UPS until manual sudo reboot
05:12:28 from device 192.168.1.142 (owner confirmation pending), bserv up 05:17:03,
VMs cascaded back 05:18-05:23 (no libvirt autostart — known gap since 04.11).
ovpn booted 05:20:04, post-boot verified healthy: openvpn-client@vpn-udp active
since 05:21:53, tun0 10.8.0.2/24, ping 10.8.0.1 ~15 ms 0% loss, nginx active,
git.gnexus.space HTTPS 200, minecraft-tcp-forward listening :25565, disk / 69%,
RAM ok, dmesg clean. Panel offline alerts for VMs (#1326-#1344, 05:23:31 UTC,
silent 37-38 min) were a delayed batch written when the panel host itself came
back.'
owner: gmikcon
purpose: VPN egress node (openvpn client, nginx vhost entry, Minecraft socat fwd)
source_of_truth: ssh-host-verified
status: active
type: vm
user: ubuntu
- id: gmikcon-pc
active_services:
- ollama.service
- sshd.service
addresses:
lan:
- 192.168.1.129
cpu_model: AMD Ryzen 5 3600 6-Core
docs: ../10-systems/hosts/gmikcon-pc.md
environment: home
gpu: NVIDIA GeForce RTX 3090 24GB (driver 615.78.08)
last_reviewed: '2026-11-08'
name: gmikcon-pc
notes: 'Личный десктоп владельца (Arch Linux, zsh, user gmikcon; sudo = SSH-пароль,
gnexus-creds secret ''gmikcon PC (192.168.1.129) — SSH gmikcon'' id cfc8ee47).
Порты 22 (SSH, есть fail2ban-подобный cooldown после неудачных попыток) и 11434
(Ollama, открыт в LAN). MAC 74:56:3c:b4:cb:18. RAM 62 GiB. 2026-11-08 оптимизация
Ollama под qwen3.6:35b-128k (архитектура qwen35moe: 256 экспертов/8 активных,
гибридное внимание full_attention_interval=4, native ctx 262144, GQA 2 KV-heads,
capabilities completion+vision+tools+thinking) под 3 параллельных запроса x 128k
ctx + 3 ГБ VRAM-резерв под ОС. Ключевой шаг: обновление Ollama 0.30.11 -> 0.40.15
(старый планировщик не давал parallel для recurrent/moe GGUF; 0.40.15 снял блок),
num_gpu удалён из params манифеста модели (блокировал fit: ''n_gpu_layers already
set by user'') — создана отдельная модель qwen3.6:35b-128k-fit (оригинал не тронут).
Итоговый systemd override /etc/systemd/system/ollama.service.d/override.conf (бэкап
override.conf.bak-rei-20261108): OLLAMA_HOST=0.0.0.0:11434, FLASH_ATTENTION=1,
KV_CACHE_TYPE=q8_0, NUM_PARALLEL=3, CONTEXT_LENGTH=131072, KEEP_ALIVE=-1, MAX_LOADED_MODELS=1,
OLLAMA_GPU_OVERHEAD=3000000000 (3 GiB резерв), DEBUG убран после отладки, LLAMA_ARG_N_CPU_MOE=19
(первые 19 из 42 слоёв MoE-эксперты в CPU/RAM, 23 последних на GPU; подбором по
расчёту: free 14572 MiB -> вернуть 10872 -> 23.5 слоя). KV q4 не понадобился:
q8_0 влез (KV-буфер 4080 MiB = 3 слота x ~1360 MiB). Итог: single-stream генерация
61.3 tok/s (было 38 в режиме все-эксперты-в-RAM), VRAM 19812/24576 MiB (свободно
~4.7 GiB, резерв 3 GiB соблюдён), RSS llama-server ~10.9 GiB RAM. Приёмочный тест:
3 одновременных запроса (ru/ru/en), все done=true, 30-38 tok/s на поток, total
wall 6.5 s, VRAM стабильна. Установка 0.40.15: ''Text file busy'' на работающих
CLI-процессах владельца (ollama launch) обойдён через mv-переименование старых
бинарей (rename работает на занятых файлах), старый бинарь сохранён для отката.
Модели на хосте: qwen3.6:35b-128k, qwen3.6:35b (22 GB), gemma4:e2b-it-qat, nomic-embed-text.'
purpose: 'Личный десктоп владельца: локальный LLM-инференс (Ollama, MoE-модели)'
source_of_truth: ssh-host-and-owner-confirmed
status: active
type: workstation
user: gmikcon