Observability Stack

The Problem — Your Server Is Down and You Have No Idea Why

Your server is down and you have no idea why. You SSH in and grep through logs — if you're lucky. If not, you restart and hope. CPU spikes, disk fills, memory leaks, crashed containers — you only find out about them when users start yelling.

Without metrics, logs, and dashboards, every outage is a fire drill. You're debugging blind, reacting instead of preventing, and spending hours on what should be 30-second investigations.

What This Stack Does For You

See CPU spikes, disk fills, and log errors on a dashboard before they become incidents. Deploy Grafana, Prometheus, Loki, and Alloy on any Linux server in minutes.

What You'll Be Able To Do After Deploying

• See your entire infrastructure on one dashboard — Pre-built Grafana dashboards show CPU, memory, disk, network, and container metrics within 30 seconds of deploying. Spot anomalies before they become outages.
• Search logs without SSH — Loki aggregates logs from every container, searchable from Grafana. No more docker logs on five different containers to find an error.
• Get alerted before users do — Prometheus auto-discovers services and collects metrics with 30-day retention. Configure alerts for disk space, memory pressure, or container restarts.
• Monitor every container automatically — cAdvisor feeds per-container resource usage into Prometheus so you see exactly which container is eating your CPU
• Track host-level health — Node Exporter reports CPU, memory, disk, and network stats from the host, not just containers
• Reload config without restarts — Prometheus lifecycle API supports configuration reloads without downtime

Why This Saves You Hours

Setting up observability from scratch means:

• Piecemeal setup: Installing Prometheus, configuring Grafana data sources, deploying cAdvisor, wiring Loki — each with its own config syntax and quirks
• Dashboard building: Creating dashboards from scratch or hunting for the right JSON on Grafana's community site
• Log pipeline: Deploying Loki, setting up log collectors, configuring label extraction — then debugging why your logs don't show up
• Missed incidents: Without dashboards, you discover problems when users report them — not when metrics start trending wrong

This stack gives you all 7 components with pre-built dashboards and auto-provisioned data sources. One docker compose up and you're observing, not guessing.

What You Get

• Grafana (v11.4) — Pre-configured with auto-provisioned data sources and dashboards
• Prometheus (v2.55) — Metrics collection with 30-day retention
• Loki (v3.2) — Log aggregation with Alloy collector
• node_exporter — Host-level metrics (CPU, memory, disk, network)
• cadvisor — Container resource monitoring

Features:

• Auto-provisioned data sources (Prometheus + Loki connected at startup)
• Pre-built dashboards: Node Exporter Full + Docker Container Monitoring
• Alloy log collector with Docker container log discovery
• Environment-based configuration (.env file)
• Prometheus lifecycle API for config reloads
• Configurable retention, ports, and version pinning

Delivery:

• docker-compose.yml
• .env.example
• README.md with architecture diagram and production checklist
• config/ directory with all service configurations and provisioning
• Pre-built Grafana dashboard JSON definitions

Requirements

• Linux server (x86_64 or ARM64)
• Docker Engine ≥ 24.x
• Docker Compose ≥ v2.24

Your Outcome

5 minutes from now, you'll have Grafana dashboards showing CPU, memory, disk, and container metrics — with Loki log aggregation and Prometheus alerting — deployed on any Linux server. No more SSH-and-grep fire drills. No more finding out about outages from users.

Note: You'll need to create a Stripe price ID in your Stripe Dashboard and update stripe_price_id in this file before purchases work.