Skip to content

DEPLOY.md

Latest commit

 

History

History
140 lines (106 loc) · 4.15 KB

DEPLOY.md

File metadata and controls

140 lines (106 loc) · 4.15 KB
Raw

Datacenter Monitoring — Deployment Guide

Quick Start (Mock Sensors)

1. Verify mock sensor works

cd skills/datacenter-monitoring
python3 scripts/mock-sensor.py --all --summary

Expected output:

🏢 Datacenter Sensor Readings — 2026-02-17 12:00:00 UTC

Rack   Temp (°C)    Humidity (%)   Status
----------------------------------------------
A1     23.8         44.5           normal
A2     24.1         45.2           normal
A3     24.3         42.8           normal
B1     22.2         46.3           normal
B2     23.1         47.1           normal
B3     25.2         41.8           normal

✅ All racks within normal range.

2. Test anomaly scenarios

# Temperature spike on rack A1
python3 scripts/mock-sensor.py --rack A1 --spike

# Sensor failure on rack B2
python3 scripts/mock-sensor.py --rack B2 --fail

# Random chaos (10% failures, 15% spikes)
python3 scripts/mock-sensor.py --all --chaos --summary

# HVAC status
python3 scripts/mock-sensor.py --hvac

3. Configure picoclaw cron

Add this to ~/.picoclaw/config.json (or use picoclaw cron add):

{
  "cron_jobs": [
    {
      "name": "dc_temp_check",
      "schedule": "*/5 * * * *",
      "command": "Read all datacenter rack temperatures by running 'python3 skills/datacenter-monitoring/scripts/mock-sensor.py --all'. Analyze the results. If any rack shows 'warning' or 'critical' status, cross-check with adjacent rack readings. If this is a confirmed hot spot, alert via the configured messaging channel. If only one sensor shows an issue, flag it as a potential sensor malfunction. Also check HVAC status with '--hvac' flag. Report findings using the report_gene tool to solidify the experience."
    }
  ]
}

4. Start the gateway

picoclaw gateway

5. Monitor gene evolution

# Check gene pool after a few hours
picoclaw gene list
picoclaw gene stats

Production Deployment (Real Sensors)

Hardware

Component Model Connection Notes
LicheeRV Nano RISC-V 1GHz — Main edge board
DS18B20 x3 1-Wire temp GPIO 4 (default) One per rack, waterproof probe
DHT22 I2C humidity GPIO 17 Ambient humidity
4ch Relay 5V relay module GPIO 18-21 Fan/HVAC control

Wiring

LicheeRV Nano
├── GPIO4  ─── DS18B20 (Rack A) ─── 4.7kΩ pullup to 3.3V
│               DS18B20 (Rack B) ─── (shared 1-Wire bus)
│               DS18B20 (Rack C) ─── (shared 1-Wire bus)
├── GPIO17 ─── DHT22 data pin ─── 10kΩ pullup to 3.3V
├── GPIO18 ─── Relay CH1 (exhaust fan)
├── GPIO19 ─── Relay CH2 (backup cooling)
├── 3.3V   ─── sensor VCC
├── 5V     ─── relay VCC
└── GND    ─── common ground

Software Setup

# Enable 1-Wire on Linux
echo "w1-gpio" | sudo tee -a /etc/modules
echo "w1-therm" | sudo tee -a /etc/modules

# Read DS18B20
cat /sys/bus/w1/devices/28-*/w1_slave

# Install picoclaw
wget https://github.com/Clawland-AI/picclaw/releases/latest/download/picclaw-linux-arm64
chmod +x picclaw-linux-arm64
sudo mv picclaw-linux-arm64 /usr/local/bin/picoclaw

# Configure
picoclaw onboard
# Edit ~/.picoclaw/config.json with your LLM API key and channel settings

Replace mock-sensor.py with read-ds18b20.py in the cron command for real sensor readings.

Gene Evolution Timeline

Period Expected Behavior
Day 1-3 Uses seed gene gene_dc_temp_spike_crosscheck at 70% confidence
Day 3-7 Confidence rises to 85%+ after successful cross-checks
Week 2 Creates time-of-day pattern genes (if afternoon temps consistently higher)
Week 3 Creates rack-specific genes (if some racks consistently hotter)
Month 2+ High-confidence genes shared to Fleet for other datacenter nodes

Cost Comparison

Item Traditional Clawland
Hardware $5,000+ monitoring system $28 sensor kit
Software $2,000/yr DCIM license Free (open source)
Labor $48,000/yr night shift operator $0 (automated)
Annual total $55,000+ $28 one-time