Prometheus — Open Source Monitoring & Alerting Toolkit

What it is

Prometheus is an open-source monitoring system and time series database, originally built at SoundCloud and now a CNCF-graduated project. It uses a pull-based model to scrape metrics from instrumented targets at configured intervals, stores them locally, and provides PromQL -- a powerful query language for aggregation, filtering, and alerting.

It is designed for DevOps engineers and SREs who need reliable metrics collection, alerting, and dashboarding for containerized and cloud-native workloads.

How it saves time or tokens

Prometheus auto-discovers scrape targets in Kubernetes using service discovery, eliminating manual target configuration as services scale up or down. PromQL lets you write complex queries -- rate of HTTP errors over 5 minutes, 99th percentile latency per endpoint -- in a single expression. The built-in Alertmanager routes alerts to Slack, PagerDuty, or email based on configurable rules, replacing custom alerting scripts.

How to use

1. Start Prometheus with Docker.

docker run -d --name prometheus -p 9090:9090 \
  -v ./prometheus.yml:/etc/prometheus/prometheus.yml \
  prom/prometheus:latest

1. Create a minimal prometheus.yml.

global:
  scrape_interval: 15s
scrape_configs:
  - job_name: prometheus
    static_configs:
      - targets: ['localhost:9090']

1. Open http://localhost:9090 and query metrics with PromQL.

rate(prometheus_http_requests_total[5m])

Example

A PromQL query for alerting on high error rates:

sum(rate(http_requests_total{status=~"5.."}[5m]))
  /
sum(rate(http_requests_total[5m]))
  > 0.05

This fires when more than 5 percent of HTTP requests return 5xx status codes over a 5-minute window.

Related on TokRepo

• Monitoring tools -- Compare Prometheus with other observability solutions.
• DevOps tools -- Infrastructure automation that pairs with monitoring.

Common pitfalls

• Prometheus stores data locally by default. For long-term retention, use Thanos or Cortex as a remote storage backend.
• High-cardinality labels (user IDs, request IDs) cause memory usage to explode. Keep label cardinality bounded.
• The pull model requires network access from Prometheus to all targets. In firewalled environments, use Pushgateway for short-lived jobs.