Uptime Kuma Monitoring MLOps Workflow — ตรวจสอบ

Uptime Kuma MLOps

Uptime Kuma Open Source Self-hosted Monitoring MLOps Workflow ML Pipeline Health Check Alert Telegram Slack Discord Status Page Docker Heartbeat Push Monitor

Installation และ Setup

=== Uptime Kuma Setup ===

Docker — Single Command

docker run -d \

--restart=always \

--name uptime-kuma \

-p 3001:3001 \

-v uptime-kuma:/app/data \

louislam/uptime-kuma:1

Docker Compose — Production

version: "3"

services:

uptime-kuma:

image: louislam/uptime-kuma:1

container_name: uptime-kuma

restart: always

ports:

• "3001:3001"

volumes:

• ./data:/app/data

environment:

• NODE_EXTRA_CA_CERTS=/app/data/certs/ca.pem

caddy:

image: caddy:2

ports:

• "80:80"
• "443:443"

volumes:

• ./Caddyfile:/etc/caddy/Caddyfile

depends_on:

• uptime-kuma

Caddyfile — Auto SSL

status.example.com {

reverse_proxy uptime-kuma:3001

}

API — Create Monitor Programmatically

import requests

KUMA_URL = "http://localhost:3001"

เนื้อหาเกี่ยวข้อง — ดูเพิ่มเติมเรื่อง LLM Fine-tuning LoRA Consensus Algorithm

TOKEN = "your-api-token"

def create_monitor(name, url, monitor_type="http"):

response = requests.post(

f"{KUMA_URL}/api/monitors",

headers={"Authorization": f"Bearer {TOKEN}"},

json={

"name": name,

แนะนำเพิ่มเติม — บทวิเคราะห์จาก XM Signal

"url": url,

"type": monitor_type,

"interval": 60,

"retryInterval": 30,

"maxretries": 3,

"accepted_statuscodes": ["200-299"],

}

)

return response.json()

from dataclasses import dataclass

@dataclass

class Monitor:

name: str

type: str

target: str

interval: int

เนื้อหาเกี่ยวข้อง — Distributed Tracing MLOps Workflow

uptime_30d: float

avg_ms: int

status: str

monitors = [

Monitor("ML API /predict", "HTTP", "https://ml.example.com/predict", 30, 99.95, 85, "Up"),

Monitor("ML API /health", "HTTP", "https://ml.example.com/health", 60, 99.99, 12, "Up"),

Monitor("Model Registry", "HTTP", "https://mlflow.example.com", 60, 99.90, 150, "Up"),

Monitor("GPU Server SSH", "TCP", "gpu-01:22", 60, 99.85, 5, "Up"),

Monitor("Training Pipeline", "Push", "Heartbeat every 1h", 3600, 98.5, 0, "Up"),

Monitor("Inference GPU", "TCP", "gpu-02:8080", 30, 99.92, 8, "Up"),

Monitor("Data Pipeline", "Push", "Heartbeat every 6h", 21600, 99.0, 0, "Up"),

]

แนะนำเพิ่มเติม — คู่มือเทรดจาก SiamCafeBook

print("=== Uptime Kuma Monitors ===")

for m in monitors:

print(f" [{m.status}] {m.name} ({m.type})")

print(f" Target: {m.target} | Interval: {m.interval}s")

print(f" Uptime: {m.uptime_30d}% | Avg: {m.avg_ms}ms")

MLOps Monitoring

=== MLOps Workflow Monitoring ===

Push Monitor — Training Pipeline Heartbeat

import requests

PUSH_URL = "https://status.example.com/api/push/abc123"

def report_training_step(epoch, loss, accuracy):

"""Send heartbeat after each training epoch"""

msg = f"Epoch {epoch}: loss={loss:.4f}, acc={accuracy:.4f}"

เนื้อหาเกี่ยวข้อง — ทำความเข้าใจ Healthchecks.io CQRS Event Sourcing

requests.get(f"{PUSH_URL}?status=up&msg={msg}&ping=")

def report_training_failed(error):

"""Report training failure"""

requests.get(f"{PUSH_URL}?status=down&msg={error}")

# In training loop:

for epoch in range(100):

loss, acc = train_epoch(model, dataloader)

report_training_step(epoch, loss, acc)

# Keyword Monitor — Check Model Version

# Monitor URL: https://ml.example.com/health

# Expected keyword: "model_version":"v2.1"

# Alert if model version changed unexpectedly

HTTP Monitor — Inference Latency

Monitor: https://ml.example.com/predict

Method: POST

Body: {"text": "test input"}

Expected status: 200

Max response time: 500ms (alert if exceeded)

@dataclass

class MLWorkflow:

stage: str

monitor_type: str

check: str

alert_condition: str

severity: str

workflows = [

เนื้อหาเกี่ยวข้อง — ทำความเข้าใจ Delta Lake — คู่มือฉบับสมบูรณ์ 2026

MLWorkflow("Data Ingestion", "Push", "Heartbeat ทุก 1h", "Missing > 2h", "Warning"),

MLWorkflow("Feature Engineering", "Push", "Heartbeat ทุก 30m", "Missing > 1h", "Warning"),

MLWorkflow("Model Training", "Push", "Heartbeat ทุก Epoch", "Missing > 2h", "Critical"),

MLWorkflow("Model Evaluation", "HTTP", "Eval endpoint", "Score < threshold", "Critical"),

MLWorkflow("Model Registry", "HTTP", "MLflow /health", "Down > 3min", "High"),

MLWorkflow("Model Serving", "HTTP", "/predict latency", "p99 > 500ms", "Critical"),

MLWorkflow("GPU Health", "TCP", "GPU server port", "Unreachable", "Critical"),

MLWorkflow("Data Quality", "Push", "DQ check ทุก 6h", "Missing > 12h", "Warning"),

]

print("\n=== MLOps Workflow Monitors ===")

for w in workflows:

print(f" [{w.severity}] {w.stage}")

print(f" Type: {w.monitor_type} | Check: {w.check}")

print(f" Alert: {w.alert_condition}")

Alert และ Status Page

# === Alert Configuration ===

# Telegram Alert
# Bot Token: from @BotFather
# Chat ID: from @userinfobot
# Setup in Uptime Kuma: Settings > Notifications > Telegram

# Line Notify
# Token: from https://notify-bot.line.me/
# Setup: Notifications > Line Notify > Token

# Discord Webhook
# URL: Server Settings > Integrations > Webhooks
# Setup: Notifications > Discord > Webhook URL

# Status Page
# Uptime Kuma > Status Pages > New
# Public URL: https://status.example.com
# Groups:
#   - ML API Services
#   - Training Pipeline
#   - Infrastructure
# Custom Domain + SSL via Caddy/Nginx

@dataclass
class AlertChannel:
    channel: str
    setup: str
    speed: str
    use_case: str
    enabled: bool

channels = [
    AlertChannel("Telegram", "Bot Token + Chat ID", "< 5s", "Primary on-call", True),
    AlertChannel("Slack", "Webhook URL", "< 5s", "Team channel", True),
    AlertChannel("Discord", "Webhook URL", "< 5s", "Dev team", True),
    AlertChannel("Line Notify", "Token", "< 10s", "Thai team", True),
    AlertChannel("Email", "SMTP config", "< 60s", "Management", True),
    AlertChannel("PagerDuty", "Integration key", "< 5s", "Critical only", False),
    AlertChannel("Webhook", "Custom URL", "< 5s", "Automation", True),
]

print("Alert Channels:")
for c in channels:
    status = "ON" if c.enabled else "OFF"
    print(f"  [{status}] {c.channel}")
    print(f"    Setup: {c.setup} | Speed: {c.speed} | Use: {c.use_case}")

# Maintenance Window
maintenance = [
    "ตั้ง Maintenance Window ก่อน Deploy ป้องกัน False Alert",
    "ใช้ API ตั้ง Maintenance อัตโนมัติจาก CI/CD",
    "Status Page แสดง Scheduled Maintenance ให้ลูกค้า",
    "Backup Data ทุกวัน /app/data/kuma.db",
    "Update Image ทุกเดือน docker pull louislam/uptime-kuma:1",
]

print(f"\n\nMaintenance Tips:")
for i, m in enumerate(maintenance, 1):
    print(f"  {i}. {m}")

เคล็ดลับ

• Docker: ติดตั้ง Docker คำสั่งเดียว ง่ายมาก
• Push: ใช้ Push Monitor สำหรับ ML Pipeline Heartbeat
• Keyword: ตรวจ Model Version ใน Health Response
• SSL: ใช้ Caddy Auto SSL สำหรับ Status Page
• Backup: Backup kuma.db ทุกวัน สำคัญมาก

การนำไปใช้งานจริงในองค์กร

สำหรับองค์กรขนาดกลางถึงใหญ่ แนะนำให้ใช้หลัก Three-Tier Architecture คือ Core Layer ที่เป็นแกนกลางของระบบ Distribution Layer ที่ทำหน้าที่กระจาย Traffic และ Access Layer ที่เชื่อมต่อกับผู้ใช้โดยตรง การแบ่ง Layer ชัดเจนช่วยให้การ Troubleshoot ง่ายขึ้นและสามารถ Scale ระบบได้ตามความต้องการ

เรื่อง Network Security ก็สำคัญไม่แพ้กัน ควรติดตั้ง Next-Generation Firewall ที่สามารถ Deep Packet Inspection ได้ ใช้ Network Segmentation แยก VLAN สำหรับแต่ละแผนก ติดตั้ง IDS/IPS เพื่อตรวจจับการโจมตี และทำ Regular Security Audit อย่างน้อยปีละ 2 ครั้ง

สรุป

Uptime Kuma Self-hosted Monitoring MLOps Workflow Push Heartbeat HTTP TCP Alert Telegram Slack Discord Line Status Page Docker Caddy SSL Production Pipeline