4 Golden Signals (Google SRE):
│
├── Latency → Πόσο αργά απαντά;
│ (p50, p95, p99 response times)
│
├── Traffic → Πόσα requests;
│ (req/sec, connections/sec)
│
├── Errors → Πόσα errors;
│ (HTTP 5xx rate, failed tasks)
│
└── Saturation → Πόσο γεμάτο;
(CPU %, RAM %, disk I/O, connections)
Για Docker containers:
├── Container: CPU, RAM, network I/O, disk I/O
├── Host: Load, disk space, network
└── App: Response time, error rate, throughput
Monitoring Stack:
┌─────────────────────────────────────────────────┐
│ Managed Hosts │
│ ┌──────────────┐ ┌──────────────┐ │
│ │ node_exporter│ │ cAdvisor │ │
│ │ :9100 │ │ :8080 │ │
│ │ (host metrics)│ │(container │ │
│ └──────┬───────┘ │ metrics) │ │
│ │ └──────┬───────┘ │
└─────────┼────────────────┼─────────────────────┘
│ Pull metrics │
▼ ▼
┌──────────────────────────────────────────────┐
│ Monitoring Server │
│ │
│ ┌─────────────┐ ┌──────────────────────┐ │
│ │ Prometheus │ │ Alertmanager │ │
│ │ :9090 │──│ :9093 │ │
│ │ (scrape + │ │ (routing alerts) │ │
│ │ storage) │ └──────────────────────┘ │
│ └──────┬──────┘ │
│ │ │
│ ┌──────▼──────┐ │
│ │ Grafana │ │
│ │ :3000 │ │
│ │ (dashboards│ │
│ └─────────────┘ │
└──────────────────────────────────────────────┘
Pull model (Prometheus):
Prometheus ──► scrape ──► node_exporter
Πλεονεκτήματα:
✅ Centralized control
✅ Εύκολο service discovery
✅ Εύκολο να δεις αν host είναι down
Push model (Telegraf, StatsD):
Container ──► push ──► InfluxDB
Πλεονεκτήματα:
✅ Firewall-friendly (outbound μόνο)
✅ Short-lived containers
→ Χρησιμοποιούμε Pull (Prometheus)
για static infrastructure
Monitoring Philosophy
│
├── 4 Golden Signals: Latency, Traffic, Errors, Saturation
├── Stack: node_exporter → Prometheus → Grafana
├── Alerts: Prometheus rules → Alertmanager → Slack/Email
└── Pull model: Prometheus scrapes exporters
