Prometheus: Self-Hosted Monitoring and Alerting

Prometheus is the industry-standard open-source monitoring and alerting toolkit. With over 62,000 GitHub stars, it is the second project to graduate from the Cloud Native Computing Foundation (after Kubernetes) and has become the backbone of observability for modern infrastructure. Prometheus collects time-series metrics through a pull-based model, stores them in a custom time-series database, and provides PromQL — a powerful query language for slicing and aggregating your data.

Self-hosting Prometheus gives you a robust monitoring foundation with complete ownership of your metrics data and alerting configuration.

Key Features

• Pull-based metrics collection via HTTP endpoints — no agents pushing data to a central server
• PromQL query language for flexible metric aggregation, filtering, and mathematical operations
• Built-in Alertmanager for routing alerts to email, Slack, PagerDuty, and webhooks
• Multi-dimensional data model with key-value label pairs on every metric
• Service discovery for automatically finding targets in Kubernetes, Consul, EC2, and Docker
• Federation support for hierarchical metric collection across multiple Prometheus instances
• Thousands of community exporters for databases, hardware, applications, and cloud services
• Efficient local storage with configurable retention and optional remote write to long-term stores

Why Self-Host Prometheus?

Metrics data stays private. Your monitoring data reveals infrastructure topology, traffic patterns, error rates, and resource utilization. Self-hosting Prometheus keeps all of this on your own servers. No third-party SaaS vendor has visibility into your operational metrics.

No per-metric pricing. Cloud monitoring services charge by the number of metrics, time-series, or data points ingested. These costs grow unpredictably as your infrastructure scales. Self-hosted Prometheus on a fixed-cost VPS ingests unlimited metrics with no surprise bills.

Complete alerting control. Alertmanager gives you full control over alert routing, grouping, silencing, and inhibition. Configure exactly which channels receive which alerts, with complex routing trees that match your team's on-call structure.

Ecosystem integration. Prometheus is the monitoring standard that almost everything integrates with. Kubernetes exposes Prometheus metrics natively. Grafana uses PromQL as a first-class query language. Thousands of exporters exist for every database, message queue, and infrastructure component.

System Requirements

Resource	Minimum	Recommended
CPU	2 vCPUs	4 vCPUs
RAM	2 GB	8 GB
Storage	20 GB SSD	100 GB SSD
OS	Ubuntu 22.04+	Ubuntu 24.04

Prometheus stores all metrics in local storage. RAM usage scales with the number of active time series (roughly 1-2 KB per series). Storage depends on retention period and scrape frequency. For a typical homelab or small production setup monitoring 50-100 targets, 4 GB RAM and 50 GB storage is comfortable.

Getting Started

Deploy Prometheus on your VPS using Docker Compose through Dokploy. Our guide covers scrape configuration, Alertmanager setup, Grafana integration, persistent storage, and SSL access.

Deploy Prometheus with Dokploy →

Alternatives

• Netdata — Real-time monitoring with zero-configuration auto-discovery and built-in dashboards
• Grafana — Visualization platform often paired with Prometheus for dashboards
• InfluxDB — Time-series database with its own query language and built-in dashboards
• VictoriaMetrics — Prometheus-compatible time-series database with better compression and performance

FAQ

What is PromQL? PromQL (Prometheus Query Language) is a functional query language for selecting, aggregating, and transforming time-series data. It lets you write queries like rate(http_requests_total[5m]) to calculate request rates, or topk(5, container_memory_usage_bytes) to find the top 5 memory consumers.

How does Prometheus differ from Grafana? Prometheus collects and stores metrics. Grafana visualizes them. They are complementary tools typically deployed together. Prometheus handles the data pipeline (scraping, storage, alerting), while Grafana provides the dashboards and visualization layer.

Can Prometheus handle large-scale monitoring? Yes, with proper architecture. A single Prometheus instance can handle millions of time series. For larger deployments, use federation to aggregate metrics from multiple Prometheus instances, or use Thanos/Cortex for horizontally scalable long-term storage.

Does Prometheus require exporters on every target? Many applications expose Prometheus metrics natively (Kubernetes, etcd, Grafana, Traefik). For applications that don't, you install an exporter — a lightweight process that translates application metrics into Prometheus format. Popular exporters exist for Node (system metrics), MySQL, PostgreSQL, Redis, and hundreds more.

---

App data sourced from selfh.st open-source directory.