Implementeaza planul claude-master-plan-discord-bridge-20260830 (15 taskuri,
3 lane-uri paralele) — un bot subtire discord.py peste CLI-ul `claude`, cu
proces persistent per fir alimentat pe stdin cu --input-format stream-json.
Nucleu: runner (proces persistent + reaper 20min + respawn --resume), stream
(parser tolerant), session_store (scriere atomica, lock per fir, detectare PID
reuse, recovery), limits (max 4 procese, timeout tur, rate per user, plafon cost
pe zi), render (un loop de editare per canal, interval adaptiv).
Adaptor: allowlist guild/canal/user fail-closed cu respingerea webhook-urilor,
comenzi !new/!cd/!model/!status/!stop/!cleanup, cost si model in subsolul
fiecarui raspuns. Mesajul sosit in timpul unui tur devine steering, nu tur nou.
Securitate: hook PreToolUse fail-closed care cere confirmare in Discord pentru
operatiuni ireversibile, wrapper `infra` cu lista explicita de hosturi. Deny
rules raman strat cosmetic, nu bariera (verificat: /usr/bin/ssh trece pe langa).
Ops: alerte email pe conventia repo-ului, !cleanup pentru orfani, unit systemd
user cu KillMode=control-group si limite de memorie, install.sh idempotent.
Verificat: 275 teste fara retea/Discord/API (10.8s), identic cu si fara
discord.py instalat; e2e pe CLI real confirma steering-ul mid-tur (mesaj la 6s
intr-un tool call de 25s schimba raspunsul final).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01B29CApsP1JkSdjYaGaHpE7
LXC 110 (moltbot) a picat dupa un OOM local containerului urmat de reboot:
tailscaled a ramas delogat, iar containerul mostenea resolv.conf-ul Tailscale
de la host-ul pve1 (doar MagicDNS 100.100.100.100) -> rezolutie DNS zero desi
L3 era functional -> echo-core in crash-loop pe telegram.error.TimedOut.
Fixuri aplicate:
- pct set 110 --nameserver '10.0.20.1 1.1.1.1' (elimina dependenta de Tailscale)
- zram-tools pe pve1 (8G zstd, prio 100) - `swap: 4096` din config era fictiv,
host-ul nu avea niciun swap; root pe ZFS deci zram, nu swapfile
- MemoryHigh/MemoryMax pe pocket-tts + supertonic-tts - toate serviciile user
rulau cu limite `infinity`, de unde OOM-uri recurente (Apr 25, May 28 x4)
cu victime aleatorii alese dupa oom_score_adj
Documentatie:
- nou post-mortem in cluster/incidents/, adaugat in indexuri
- README lxc110: host corectat pveelite -> pve1 (+ RAM/CPU/storage reale),
comenzile pct redirectionate spre nodul corect
- README lxc171: tabel cu starea swap pe cele 3 noduri (pveelite are zvol,
nu zram - nu necesita acelasi fix)
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Incident 2026-06-24: OOM-uri repetate în cgroup /lxc/171 cauzate de swap
nebacked pe host (pvemini fără swap) + acumulare de forks vscode-server
orfane și sesiuni logind zombie.
- scripts/reap-orphans.sh: reaping conservator (forks ~/.vscode-server
orfane >24h + sesiuni closing cu leader mort), rulat din cron la 6h
- README: secțiune Memorie & OOM (zram pe host ZFS, reaper, diagnostic),
corectat date stale host/RAM/CPU (pvemini, 16GB, 4 cores)
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
- Rename proxmox/claude-agent/ to proxmox/lxc171-claude-agent/
- Move scripts to scripts/ subdirectory
- Add complete installation guide for new LXC from scratch
- Update proxmox/README.md with LXC 171 documentation and navigation
- Add LXC 171 to containers table
- Remove .serena/project.yml
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
