Skip to content

Commit 4497a0b

Browse files
leopu00claude
andcommitted
docs(sessions): VPS+container health audit (snapshot 14:09 UTC)
Audit operativo Path 2 VPS Hetzner durante finestra Kimi 5 in corso. Verifica: ordine agenti, aderenza istruzioni, dipendenze, salute infra, riempimento disk, posizionamento file. Risultato 8/10: - 10/11 agenti attivi (manca Dottore — conferma bug #18) - File structure ordinata: /jht_user/{cv,critiche,allegati,output} - Worktrees puliti, 0 file mal-piazzati in agents/ - Disk 14% (10G/75G), RAM 56%, swap leggero - 5 bridge attivi (sentinel, pacing, 3x tg) - Compliance envelope buona (29 ACK su 100 msgs) 3 micro-debiti tecnici emersi: - /tmp/ 24M con 221 PNG/HTML (migrare a /jht_user/output/charts/) - Cache uv 270M crescente, nessuna auto-prune (Dottore mancante) - Proj 161% cold-start bias (finestra 10 min, bug #5) Conferme empiriche di 4 bug strategici: #18, #16, #5, #2. Tempo totale audit: ~30s in 2 round paralleli SSH. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
1 parent e2558c2 commit 4497a0b

1 file changed

Lines changed: 156 additions & 0 deletions

File tree

Lines changed: 156 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,156 @@
1+
# 2026-05-17 — Health audit VPS + container (snapshot 14:09 UTC)
2+
3+
Audit operativo richiesto durante la sessione test e2e Path 2. Verifica:
4+
ordine agenti, aderenza istruzioni, gestione dipendenze, salute VPS,
5+
riempimento disk, posizionamento file. Snapshot puntuale 14:09 UTC del
6+
17 maggio 2026 (finestra Kimi 5 in corso, post-RESET 13:11).
7+
8+
## 🏥 Salute infrastruttura
9+
10+
| Livello | Indicatore | Valore | Stato |
11+
|---|---|---|---|
12+
| 🖥️ **VPS Host** | uptime | 21h 36m | 🟢 |
13+
| | load avg (1/5/15 min) | 2.05 / 2.14 / 1.68 | 🟡 ~50% (4 core) |
14+
| | RAM | 2.1G/3.7G (56%) | 🟢 |
15+
| | swap | 305M/2.0G (15%) | 🟢 leggero |
16+
| | disk `/` | **10G/75G (14%)** | 🟢 spazio abbondante |
17+
| 🐳 **Container `jht`** | CPU | 128% (≈ 1.3 core) | 🟢 normale per 10 agenti |
18+
| | RAM | 1.8G/3.7G (48%) | 🟢 |
19+
| | disk visibile | 10G/75G (14%) | 🟢 |
20+
21+
## 🧑‍🤝‍🧑 Agenti attivi (tmux sessions)
22+
23+
| Agente | Pane creato | CPU | RAM | Stato |
24+
|---|---|---|---|---|
25+
| 🕵️ SCOUT-1 | 10:28 | 4.2% | 185M | 🟢 ACTIVE |
26+
| 🕵️ SCOUT-2 | 04:30 | 8.5% | 181M | 🟢 ACTIVE |
27+
| 🔬 ANALISTA-1 | 03:30 | 8.1% | 180M | 🟢 ACTIVE |
28+
| 🎯 SCORER-1 | 03:44 | 1.6% | 170M | 🟡 standby |
29+
| ✍️ SCRITTORE-1 | 02:47 | **15.2%** | 184M | 🟢 lavora! |
30+
| ⚖️ CRITICO-S1 | 14:03 | 5.0% | 150M | 🟢 just spawn |
31+
| 💂 SENTINELLA | 16/05 | 0.9% | 153M | 🟢 monitor |
32+
| 👨‍✈️ CAPITANO | 16/05 | 0.9% | 172M | 🟢 coordina |
33+
| 💬 ASSISTENTE | 16/05 | 3.8% | 192M | 🟢 |
34+
| 🧙‍♂️ MENTOR | 16/05 | 2.2% | 161M | 🟢 |
35+
| 🩺 **DOTTORE** |||| **❌ MAI SPAWNATO** ([bug #18](../../internal/2026-05-17-team-strategy-bugs.md#-18-dottore-mai-spawnato--watchdog-non-lo-include-bridge-mancante)) |
36+
37+
**10/11 agenti attivi**. Tutti i ruoli operativi presenti. Solo Dottore assente.
38+
39+
## 🌉 Bridge attivi
40+
41+
| Bridge | PID | Cadenza | Stato |
42+
|---|---|---|---|
43+
| 📡 sentinel-bridge | 23401 | 3 min | 🟢 ATTENZIONE proj=161% usage=31% |
44+
| ⏱️ pacing-bridge | 23417 | 15 min | 🟢 |
45+
| 📲 tg-bridge × 3 | (3 PID) | long-poll | 🟢 assistente, capitano, mentor |
46+
47+
## 📁 File structure — quanto sono ordinati
48+
49+
| Posto | Size | Stato | Note |
50+
|---|---|---|---|
51+
| `/jht_user/` | **1.4M** | 🟢 ordinato | 4 subdir corrette: `cv/`, `critiche/`, `allegati/`, `output/` |
52+
| `/jht_user/critiche/` | (md) | 🟢 | 30+ review-* in formato corretto |
53+
| `/jht_user/cv/` | (pdf) | 🟢 | 27 PDF/docx/tex deliverable |
54+
| `/jht_home/agents/` | (12 worktrees) | 🟢 | una per agente attivo + critico/scorer-2 |
55+
| `/jht_home/.cache/uv` | **270M** | 🟡 cresce | cache pip/uv normale, ma tendenza ↑ |
56+
| `/jht_home/.kimi/sessions` | 71M | 🟢 | 13 dir sessions kimi |
57+
| `/tmp/` | **24M, 221 file** | 🟡 affollato | PNG grafici (20+) + HTML scraping + tar maps |
58+
| `/jht_home/logs/` | (vario) | 🟢 | sentinel-data.jsonl, pacing, watchdog |
59+
| File mal-piazzati in `/jht_home/agents/` | 0 | 🟢 ✅ | nessun PNG/PDF errante negli worktrees |
60+
61+
**Ordine generale**: 🟢 BUONO. Critiche-CV ben separati. Worktrees
62+
puliti. Solo `/tmp` accumula PNG (problema collegato a
63+
[bug #16](../../internal/2026-05-17-team-strategy-bugs.md#-16-auto-report-periodici--auto-grafici-via-bridge-orders-feature-mancante)
64+
— i grafici dovrebbero migrare in `/jht_user/output/charts/`).
65+
66+
## 📦 Dipendenze Python
67+
68+
| Cache | Size | Stato |
69+
|---|---|---|
70+
| `/jht_home/.cache/uv` | 270M | 🟢 gestita da `uv pip install --user` (regola T13) |
71+
| `/jht_home/.cache/fontconfig` | piccola | 🟢 |
72+
73+
**Pattern corretto**: gli agenti usano `uv pip install --user`,
74+
niente pollution sistema. Eccezione: niente `pyproject.toml`
75+
`requirements.txt` in `/jht_home/` (deps ad-hoc per agente — è ok per
76+
ruoli con dipendenze diverse).
77+
78+
## 📊 Skill compliance (ultime 100 messaggi inter-agente)
79+
80+
```
81+
ACK 29 ████████████████████████████████ buon ratio risposta
82+
PACING 23 █████████████████████████ tick regolari ✅
83+
INFO 23 █████████████████████████
84+
REQ 21 ███████████████████████
85+
TG 20 ██████████████████████ interazione utente
86+
RES 19 █████████████████████
87+
REPORT 19 █████████████████████
88+
TICK 15 ████████████████
89+
MSG 14 ███████████████
90+
URG 7 ████████ ⚠️ 7% stress moments
91+
EMERGENZA 2 ███ ⚠️ 2 freeze ordinati
92+
```
93+
94+
→ Mix di tipi rispetta envelope `[@from -> @to] [TIPO]`. 🟢
95+
compliance buona, ma 9 episodi URG+EMERGENZA = sintomo Sentinella
96+
aggressiva ([bug #2](../../internal/2026-05-17-team-strategy-bugs.md#-2-sentinella-ipersensibile-freeze-totale--kill-invece-di-throttle-progressivo)).
97+
98+
## ⚠️ Anomalie rilevate
99+
100+
| # | Cosa | Severità | Bug doc |
101+
|---|---|---|---|
102+
| 1 | DOTTORE mai spawnato | 🟠 | [#18](../../internal/2026-05-17-team-strategy-bugs.md#-18-dottore-mai-spawnato--watchdog-non-lo-include-bridge-mancante) |
103+
| 2 | `/tmp/` 24M con 221 PNG/HTML | 🟡 | da migrare a `/jht_user/output/` |
104+
| 3 | Proj 161% (ATTENZIONE) ma usage 31% | 🟡 | finestra giovane (10 min), bias cold-start [#5](../../internal/2026-05-17-team-strategy-bugs.md#-5-bridge-latency-al-boot-non-ha-ancora-emesso-il-primo-campione) |
105+
| 4 | 1 errore agent-watchdog | 🟡 | "0 avviati, 0 già attivi, 1 errori" — investigare |
106+
| 5 | Cache uv 270M crescente | 🟡 | nessuna auto-prune (compito sarebbe del Dottore [#18](../../internal/2026-05-17-team-strategy-bugs.md#-18-dottore-mai-spawnato--watchdog-non-lo-include-bridge-mancante)) |
107+
108+
## 🎯 Verdict generale: 🟢 **8/10**
109+
110+
| Dimensione | Voto | Nota |
111+
|---|---|---|
112+
| 🏥 Salute infra | 9/10 | tanto spazio, RAM ok, load gestibile |
113+
| 🧑‍🤝‍🧑 Agenti ordinati | 8/10 | 10/11 attivi, file in workdir corretti, Dottore assente |
114+
| 📜 Aderenza istruzioni | 7/10 | envelope corretto ma 7 URG + 2 EMERGENZA |
115+
| 📦 Dipendenze | 9/10 | `uv pip --user` ovunque, niente leak |
116+
| 💾 Riempimento | 9/10 | 14% disk, ma cache 270M crescente |
117+
| 📂 Posizionamento file | 8/10 | `/jht_user/{cv,critiche,allegati,output}` ✅, ma 24M PNG su `/tmp/` |
118+
119+
**TL;DR**: 🚀 Team funzionante e operativo, struttura ordinata, infra
120+
serena. **3 micro-debiti tecnici**: spawn Dottore (#18), prune
121+
cache+`/tmp` (servirebbe il Dottore), bias cold-start projection (#5).
122+
Niente di urgente.
123+
124+
## 🔬 Comandi usati per l'audit
125+
126+
Per ripetere l'audit (esegui da workstation con SSH key VPS):
127+
128+
```bash
129+
ssh -i "$SSH_KEY" root@<VPS_IP> '
130+
uptime; free -h; df -h / # host
131+
docker stats jht --no-stream # container
132+
docker exec jht tmux list-sessions # agenti
133+
docker exec jht ps aux --sort=-%cpu | head -15 # CPU
134+
docker exec jht du -sh /jht_user /jht_home/.cache /tmp # disk
135+
docker exec jht find /jht_user -type f \( -name "*.pdf" -o -name "*.md" \) | wc -l # deliverables
136+
docker exec jht tail -200 /jht_home/logs/messages.jsonl | grep -oE "\"type\": \"[A-Z]*\"" | sort | uniq -c | sort -rn # compliance
137+
docker exec jht cat /jht_home/logs/sentinel-bridge-state.json
138+
'
139+
```
140+
141+
Tempo totale: ~30 secondi (2 round paralleli).
142+
143+
## 🔗 Connessioni con altri documenti
144+
145+
- [`docs/internal/2026-05-17-team-strategy-bugs.md`](../../internal/2026-05-17-team-strategy-bugs.md)
146+
— 19 bug strategici documentati. Questo audit conferma empiricamente
147+
#18 (Dottore assente), #16 (PNG su /tmp non migrati), #5 (cold-start
148+
proj), #2 (Sentinella aggressiva con 9 URG+EMERGENZA).
149+
- [`docs/sessions/2026-05-17-team-dashboard/`](../2026-05-17-team-dashboard/)
150+
`pipeline_overview.png` è il template visivo di questa stessa
151+
ispezione, ma generato dal Capitano on-demand. Se il Capitano avesse
152+
la regola C-04 ([bug #16](../../internal/2026-05-17-team-strategy-bugs.md#-16-auto-report-periodici--auto-grafici-via-bridge-orders-feature-mancante))
153+
questo audit potrebbe essere automatico ogni 2h.
154+
- [`docs/sessions/2026-05-17-budget-windows/`](../2026-05-17-budget-windows/),
155+
[`docs/sessions/2026-05-17-pipeline-snapshot/`](../2026-05-17-pipeline-snapshot/)
156+
— altre sessioni grafiche del 17 mag.

0 commit comments

Comments
 (0)