Prometheus — Introduction — Part 1

In the first part of the series we will look Prometheus introduction, features and its components

Prometheus is the de-facto standard for monitoring cloud native applications and infrastructure.

Introduction

• Prometheus is an opensource monitoring and alerting tool
• Originally built by SoundCloud in 2012
• Donated to Cloud Native Computing Foundation in 2016
• Second project hosted by CNCF after Kubernetes
• CNCF graduated project
• Collects and stores its metrics as time series data
• Metrics information is stored with the timestamp at which it was recorded
• Optional key-value pair called labels are also stored

Features

• Multi-dimensional data model — allows to define and store time-series data with multiple dimensions
• PromQL — a query language to perform complex queries and calculations on metrics data
• Time-series collection — can collect metrics data from a wide range of sources, including servers, applications, and databases
• Data retention — can set retention policies on the metrics data
• Service discovery — can automatically discover and monitor new services in a dynamic infrastructure
• Exporters — different types of agents available for collecting metrics from different types of sources
• Grafana integration — can easily integrated with Grafana for visualization and custom dashboards
• Alerting — can alert stakeholders when issue occurs

Architecture

Image source — Prometheus Documentation

Prometheus Server

• Main component of Prometheus
• Collects, stores and serve metrics data
• Follows a pull-based model for collecting metric data
• Periodically queries the targets configured in the scrape configuration
• Uses HTTP or HTTPS protocol
• Stores the data in a time-series database
• Time-series database allows Prometheus to do operations like querying, aggregation and visualization

Exporters

• Agents that collect and expose metrics data from applications or third-party systems
• Written in a wide variety of programming languages
• Designed to be lightweight and easily to deploy
• Scrape metrics data from a target and expose it on a HTTP endpoint
• Metrics data is returned in a text-based format such as plain text, JSON, or protobuf
• Official exporters are maintained by Prometheus GitHub Organization like Node Exporter, MySQL Exporter etc
• Unofficial exporters are externally contributed and maintained like Apache Exporter, PostgreSQL Exporter etc

Alertmanager

• Manages alerts generated by Prometheus
• Receives alerts from Prometheus and groups them based on criteria like severity
• Based on routing rules, it sends notification to various receivers like Email, Slack, Pagerduty etc
• Applies a set of deduplication rules to ensure that alerts are not sent multiple times for the same issue
• Provides web interface that allows users to view and manage alerts

Pushgateway

• Prometheus follows a pull-based model for collecting metrics data
• Sometimes we need to push metrics to Prometheus which cannot be scraped
• Helps to push metrics to Prometheus instead of pulling metric
• Best suitable for short-lived jobs
• Provides a HTTP API for pushing metrics from short-lived jobs

Service Discovery

• Process of identifying and monitoring systems automatically
• Helps Prometheus to keep track of different services which are running
• Supports variety of service discovery options for discovering scrape targets like Kubernetes, Consul, Docker etc
• If you need a service discovery system that currently not supported, we can use file-based service discovery
• Enables you to list scrape targets in a JSON file along with metadata
• Metadata is piece of useful information about targets, like name of the service, description etc
• HTTP service discovery helps to discover targets over an HTTP endpoint an alternative to file-based service discovery

Client Libraries

• Set of libraries that allows developers to instrument their applications to expose metrics to Prometheus
• It exposes an HTTP endpoint and Prometheus can scrape metrics from these HTTP endpoints and store them in time-series database
• Once the metrics are stored, Prometheus can generate graphs, reports and alerts based on the data
• Helps developers to get insights about their application performance
• Official client libraries are available in many programming languages like Go, Python, Java etc
• Unofficial client libraries are also available for C, C++, Node.js etc

Reference

https://prometheus.io/docs/introduction/overview/