Prometheus

• Launched by SoundCloud in 2012
• The de facto standard data model for metrics
• It's a time-series database (TSDB) + a scraper + a query language, all in one binary
• Uses a pull-model (scraping)

Components

• Retrieval
• Worker that pulls the metrics data
• Pulls from HTTP endpoints
• Default endpoint for the service: {host}/metrics
• The target must expose this /metrics route
• The metrics must be exposed in the correct format
• For services that do not expose /metrics by default it needs an exporter (to format and expose the metrics)

• For short-lived jobs the application push then into a pushgateway and prometheus scrapes from the pushgateway

• Storage

• Prometheus is deliberately simples about storage
• A single prometheus server
• Stores data on local disk (not clustered, not replicated)
• Keeps it for a limited retention (default 15 days)
  • Can also integrate with Remote Storage Systems
• Scales up but NOT out

• Due to these metrics, other solutions are used for TSDB (e.g., vicmetrics)

• HTTP server

• Accept PromSQL queries to consult the storage
• Prometheus exposes its own /metrics endpoint