P
Prometheus
Open-source metrics and monitoring — pull-based scraping, PromQL query language, native Kubernetes integration.
Visit website
👁 9784 views
About Prometheus
Prometheus is the open-source monitoring system that became the standard for Kubernetes metrics collection. It scrapes metrics from HTTP endpoints at configurable intervals and stores them as time-series data queryable with PromQL.
How Prometheus works: every application exposes a /metrics endpoint. Prometheus scrapes every configured endpoint every 15 seconds, stores the time-series, and evaluates alerting rules.
Prometheus + Grafana stack: Prometheus (metrics) + Grafana (dashboards) + Alertmanager (notifications) + Loki (logs) replaces Datadog at approximately $30/month vs $150-500/month for the same team — a 750x cost reduction.
Kubernetes-native: Prometheus has a Kubernetes discovery mode that automatically finds pods, services, and nodes to scrape. The kube-prometheus-stack Helm chart deploys Prometheus, Grafana, Alertmanager, and all Kubernetes dashboards in one command.
Common PromQL queries: rate(http_requests_total[5m]) gives requests per second. histogram_quantile(0.99, ...) gives p99 latency. sum by (service) (rate(errors_total[5m])) gives error rate per service.
Key Features
-
●
Pull-based scraping: Prometheus requests /metrics from each target at configurable intervals
-
●
Kubernetes service discovery: automatically scrapes pods, services, and nodes
-
●
PromQL: powerful query language for aggregations, rates, histograms, and label matching
-
●
Alertmanager: route alerts to Slack, PagerDuty, or email based on labels and severity
-
●
Recording rules: pre-compute expensive queries and store as new time-series for dashboard performance
Pros
- ✓The standard metrics collection layer for Kubernetes — every k8s dashboard and alert uses Prometheus format
- ✓Pull-based model means applications don''t need to push metrics — safer for firewall traversal
- ✓PromQL is the most powerful time-series query language — rate, histogram_quantile, and label matching
- ✓Native Kubernetes service discovery automatically finds pods and services without manual configuration
- ✓Apache 2.0 open-source with no enterprise tier needed for full functionality
Cons
- ✗No long-term storage by default — requires Thanos or Cortex for multi-month retention
- ✗Pull model can miss metrics for very short-lived jobs — use Pushgateway for batch jobs
- ✗PromQL has a steep learning curve compared to SQL-like query interfaces
Who is using Prometheus?
-
●
DevOps engineers running Kubernetes who need the standard metrics stack
-
●
Platform teams building observability infrastructure with the Prometheus + Grafana combination
-
●
Engineers who want Datadog-quality metrics without Datadog''s pricing
-
●
Teams adopting the kube-prometheus-stack Helm chart for complete Kubernetes observability
Use Cases
- →Scraping HTTP endpoint metrics from all microservices with Kubernetes service discovery
- →Setting up alerts for CPU usage above 80%, error rate above 1%, and p99 latency above 500ms
- →Using PromQL to calculate the 95th percentile request latency across all services
- →Storing metrics locally with 30-day retention and using Thanos for 1-year historical queries
Pricing
-
●
Open Source : $0/forever — Full monitoring, Apache 2.0, All features, Community support
Pricing details may not be up to date. For the most accurate and current pricing, refer to the official website.
What Makes Prometheus Unique?
The open-source metrics monitoring system that became the Kubernetes standard — pull-based scraping from any /metrics endpoint, PromQL for powerful time-series analysis, and native k8s discovery, replacing Datadog at 750x lower cost.
How We Rated It
Cost comparison from Datadog pricing vs self-hosted Prometheus stack on Hetzner. Kubernetes adoption data from CNCF Prometheus adopter survey.
-
Accuracy and Reliability 4.7/5
-
Ease of Use 4.2/5
-
Functionality and Features 4.8/5
-
Performance and Speed 4.8/5
-
Customer Support 4.5/5
-
Value for Money 5.0/5
AI summary
Open-source metrics and monitoring — pull-based scraping, PromQL query language, native Kubernetes integration.