CHAOS DDI ENGINE — Fleet Drift Immune System

Engines that get better as you use them.
Every fleet drift event — WSL IP change, iphlpsvc death, portproxy stale, SSH unreachable, GPU pool leak, NAS unmounted — is a MECidine signal. The engine learns from every event and every resolution. Forge's iphlpsvc crash today became 3 resolved events and a prediction model: WSL_IP_DRIFT → PORTPROXY_STALE → SERVICE_DEATH is forge's cascade pattern. Now we watch for it automatically.

DDI Bonixer + VIZASL + Minebench + MECidines. All wired. All learning. All sovereign.

chaos_ddi_engine.py forge-portproxy-refresh.ps1 6 silos tracked forge: 3 events resolved MAC pipeline integrated

🌐 FLEET CHAOS VIZASL — Live Silo Health

MSI01

192.168.2.18 · :22 · ubu-cap · LOCAL

IMMUNE STRENGTH0.500

CHAOS SCORE0.000

MINEBENCH: 0.800 · 0 events · CALM

Anchor silo — all other silos calibrate against msi01

MSCLO

192.168.2.19 · :22 · ubu-cap · RTX 5090

IMMUNE STRENGTH0.500

CHAOS SCORE0.000

MINEBENCH: 0.800 · 0 events · CALM

Known: DNS_BROKEN risk (router blocks external DNS) — fix: chattr resolv.conf

YONE

192.168.2.23 · :2222 · ubu-cap · RTX 5080

IMMUNE STRENGTH0.500

CHAOS SCORE0.000

MINEBENCH: 0.800 · 0 events · CALM

PEMCLAU prod silo — MAC pipeline primary target

PCDEV

192.168.2.16 · :22 · ubu-cap · RTX 5090 32GB

IMMUNE STRENGTH0.500

CHAOS SCORE0.000

MINEBENCH: 0.800 · 0 events · CALM

joffe-math :9384/:9385 · Watch: NAS_UNMOUNTED risk (no NAS mount visible)

FORGE / LIANLI01

192.168.2.12 · :2223 · lianli · RTX 4090

IMMUNE STRENGTH0.690 ↑

CHAOS SCORE1.000 → 0.000 ✓ RESOLVED

MINEBENCH: 0.450 · 3 events (all resolved) · RECOVERING

⚠ HIGH RISK: WSL_IP_DRIFT → PORTPROXY_STALE → SERVICE_DEATH cascade. portproxy-refresh.ps1 deployed.

EOSE-DEV

192.168.2.21 · :22 · eose · k3s cluster

IMMUNE STRENGTH0.500

CHAOS SCORE0.000

MINEBENCH: 0.800 · 0 events · CALM

SET/SET-OPS gate stack (Gatekeeper/Kyverno/Falco/Trivy) — HEARTBEAT monitored

🔬 FORGE FIX — 3 Resolved Events (Day 98)

🔴

CE-FORGE-001 · SSH_UNREACHABLE · MECidine: 0.81

forge port 2223 closed from msi01 — MAC deployment to forge blocked. Discovered during MAC fleet rollout.

✓ RESOLVED: portproxy 2223 added, iphlpsvc restarted -Force, firewall rule added

💀

CE-FORGE-002 · SERVICE_DEATH · MECidine: 0.85

iphlpsvc (IP Helper) Status=Stopped. All 39 portproxy entries silently dead. Root cause: service was never set to Automatic, stopped at some point, portproxy never worked again.

✓ RESOLVED: Restart-Service iphlpsvc -Force + Set-Service iphlpsvc -StartupType Automatic

🌀

CE-FORGE-003 · PORTPROXY_STALE · MECidine: 0.90

All 39 portproxy entries (Ollama :11434, gateways :18792/:18832, PEMOS services :9334-:9451, SSH :22/:2222/:2223) unreachable from LAN. WSL still running, sshd still listening on :2223 inside WSL — but Windows never forwarded it.

✓ RESOLVED: forge-portproxy-refresh.ps1 deployed — auto-refreshes all 39 entries on WSL restart

🧬 13 CHAOS CLASSES — MECidine Immune Training Set

WSL_IP_DRIFT

WSL IP Changes on Restart

Severity: 0.65

WSL gets new internal IP on every restart. All portproxy entries point to old IP → all services dead until refresh.

forge cascade: WSL_IP_DRIFT → PORTPROXY_STALE → SERVICE_DEATH

SERVICE_DEATH

Critical Windows/WSL Service Stopped

Severity: 0.85

iphlpsvc, sshd, docker, ollama stop unexpectedly. Portproxy dies silently when iphlpsvc stops.

Fix: Set StartupType=Automatic for all critical services

PORTPROXY_STALE

Portproxy Entry Points to Dead IP

Severity: 0.70

netsh portproxy entries survive reboots but WSL IP changes. All entries silently wrong until refresh.

Fix: forge-portproxy-refresh.ps1 at Task Scheduler logon

SSH_UNREACHABLE

SSH Port Closed / Connection Refused

Severity: 0.90

Can't connect to silo. Blocks MAC deployment, fleet-sync, cross-silo file transfer.

HEARTBEAT watchdog: nc -zw2 $ip $port before any SSH op

GPU_POOL_LEAK

AKS GPU Nodepool Count > 0

Severity: 0.95

GPU nodes left running (H100: ~CA$15/hr, T4: ~CA$1.50/hr). HEARTBEAT checks every 4 hours.

Alert Kay immediately — do NOT auto-scale-down

NAS_UNMOUNTED

NAS Mount Gone

Severity: 0.75

Cross-silo file transfer via NAS breaks silently. Affects all fleet-sync ops, MAC deployment, Lean corpus.

Primary: /mnt/nas-diskpool · msclo: /mnt/deseof · pool2: /mnt/nas-pool2

DNS_BROKEN

DNS Resolution Failure

Severity: 0.70

msclo pattern: router blocks external DNS. Fix: chattr +i /etc/resolv.conf with nameserver 100.100.100.100

Use --resolve host:443:IP with curl to bypass when broken

PEMLAAM_VIOLATION

Raw Secret Attempts Graph Admission

Severity: 0.99 🔥

Highest severity. KCF-ADA-029. If raw secret enters PEMCLAU, the vault becomes a diary with passwords.

OPA gate + entropy scanner. FULL_FORBIDDEN = quarantine immediately.

CONTAINER_CRASH

Docker Container Exits Unexpectedly

Severity: 0.80

Portal, PEMCLAU, gateway containers exit. pemos-portal watchdog checks every 15min.

restart:unless-stopped on all critical containers

BELT64_FAIL

Belt64 Round-Trip Fails

Severity: 0.75

MAC Stage 3 failure. Chunk cannot cross language/runtime boundary safely. Blocks MELIBRIX staging.

MAC bonixer auto-detects. Status=FAILED, chunk stays in FC queue.

LINEAGE_BREAK

Lineage Chain Broken in MAC Pipeline

Severity: 0.80

v1→v2→v3 transform chain loses a link. Object cannot be verified as the same organism it started as.

MAC Stage 3 lineage proof validates chain before weave.

NOVELTY_BLOCKED

MAC FC Queue Full / Novelty Below Threshold

Severity: 0.40

Low novelty chunks accumulate in FC queue. Not failures — compost waiting for context. Monitor queue depth.

HEARTBEAT: fc-flush.py --check 4x/day. Re-evaluate on new schema/entity.

PORT_CONFLICT

Two Portproxy Entries on Same Listen Port

Severity: 0.60

Seen in forge: duplicate 2223 entry added twice. Last write wins — first entry silently ignored.

portproxy-refresh.ps1: delete before add — prevents duplicates.

📊 MINEBENCH — Silo Chaos Immunity Scores

Higher = better immune system. Score = immune_strength − (chaos_score × 0.4) + (resilience × 0.3). Improves as events get resolved and engine learns new patterns.

MSI01

0.800

0 events · 0.50 immune · anchor silo

MSCLO

0.800

0 events · 0.50 immune · DNS watch

YONE

0.800

0 events · 0.50 immune · PEMCLAU prod

PCDEV

0.800

0 events · 0.50 immune · joffe-math

EOSE-DEV

0.800

0 events · 0.50 immune · k3s gates

FORGE

0.450

3 events resolved · 0.69 immune ↑ · portproxy fixed

As forge resolves more events and builds pattern history, its immune_strength will rise and minebench score will recover toward 0.80+. That's the engine learning.

🔧 FORGE STARTUP SCRIPT — Auto-Refresh on WSL Restart

forge-portproxy-refresh.ps1 — deploy to forge Windows Task Scheduler

# Task Scheduler: At logon · Run as Admin · "Run whether user is logged in or not"
# Ensures iphlpsvc running + refreshes all 39 portproxy entries with current WSL IP

Restart-Service iphlpsvc -Force -ErrorAction SilentlyContinue
$wslip = (wsl -d Ubuntu -- bash -lc "ip -4 addr show eth0 | grep -oP '(?<=inet\s)\d+(\.\d+){3}'").Trim()

# Delete and re-add all SSH + service portproxy entries with fresh WSL IP
netsh interface portproxy delete v4tov4 listenport=2223 listenaddress=0.0.0.0
netsh interface portproxy add v4tov4 listenport=2223 listenaddress=0.0.0.0 connectport=2223 connectaddress=$wslip
# ... (full version: refreshes all 39 entries, logs to ~/.forge-portproxy.log)

Test-NetConnection -ComputerName 127.0.0.1 -Port 2223 -WarningAction SilentlyContinue

Full script: fleet-sync/mac/forge-portproxy-refresh.ps1 · Copy to forge Windows and add to Task Scheduler. Logs to %USERPROFILE%\.forge-portproxy.log.

⚠ Also run when: WSL restarts · Machine reboots · If any forge service becomes unreachable from LAN

🌐 FLEET CHAOS VIZASL — Live Silo Health

🔬 FORGE FIX — 3 Resolved Events (Day 98)

🧬 13 CHAOS CLASSES — MECidine Immune Training Set

📊 MINEBENCH — Silo Chaos Immunity Scores

🔧 FORGE STARTUP SCRIPT — Auto-Refresh on WSL Restart

🔗 LINKS