Skip to main content

Metrics

Prometheus​

  • Store all data as time series

  • Group labeled data into a metric

api_http_requests_total{method="POST", handler="/messages"}

  • Collecting metrics by scraping HTTP endpoints of the services

  • Each service has to collect the metric on its behalf (by prometheus sdk, ...) and expose them following the API

  • Recommended stack: Prometheus + Grafana (for UI)

Telegraf​

  • Like Prometheus, it collects metrics by scraping HTTP endpoints

  • It often outputs to InfluxDB

  • Recommended stack: Telegraf + InfluxDB + Grafana

Calculating metrics​

  • The service counts its metric, and expose them to an HTTP Endpoint

eg: https://github.com/siimon/prom-client

  • Use log aggregator

eg: Fluentd