Card, PayPal or bank transfer · Morocco and worldwideتعلَّم · مارِس · أتقِن

Catalogue / Infrastructure & systems

High availability: HAProxy, Keepalived, Pacemaker and Corosync

Keep a service online when a server fails — load balancing with HAProxy and Nginx, a virtual IP with Keepalived, a Pacemaker/Corosync cluster with fencing, data replication, tested failovers and maintenance without downtime.

  • Intermediate to advanced
  • 12 lessons
  • 11 h 5 min
  • Versions Debian 12, Rocky 9, Ubuntu 24.04, HAProxy 2.4 (Rocky 9), 2.6 (Debian 12), 2.8 (Ubuntu 24.04), 3.0 (Debian 13), Nginx 1.20 (Rocky 9), 1.22 (Debian 12), 1.24 (Ubuntu 24.04), 1.26 (Debian 13), Keepalived 2.2 (Rocky 9, Debian 12, Ubuntu 24.04), 2.3 (Debian 13), Pacemaker 2.1 / Corosync 3.1 (Rocky 9, Debian 12, Ubuntu 24.04), Pacemaker 2.1, pcs 0.11 (Rocky 9), pcs 0.11 / crmsh 4.6 (Debian 12, Ubuntu 24.04), DRBD 9.2 (Debian 12, Rocky 9), PostgreSQL 15 (Debian 12), 16 (Ubuntu 24.04), 13-16 (Rocky 9), MariaDB 10.11 (Debian 12, Rocky 9), HAProxy 2.4-3.0, Keepalived 2.2, Pacemaker 2.1, HAProxy 2.6, DRBD 9

A MasterLinux course · Original text

What you will learn

  • Put numbers on availability (SLA, MTTR, RTO, RPO) and find a platform's single points of failure
  • Write an HAProxy configuration in L4 and L7 (frontends, backends, ACLs) and validate it with haproxy -c
  • Tune reliable health checks, session affinity and the right balancing algorithm for each service
  • Terminate TLS on HAProxy, publish the statistics page and drive servers through the admin socket
  • Use Nginx as a load balancer (upstream, passive failures, retrying on another server, TCP streams)
  • Hold a virtual IP with Keepalived/VRRP, tracked by a health script
  • Build a Pacemaker/Corosync cluster with quorum, fencing (STONITH), resources and constraints
  • Replicate data with DRBD, PostgreSQL and MariaDB, and read the state of a replication
  • Prepare and measure a failover, and carry out maintenance without downtime
  • Diagnose the major high-availability incidents (split-brain, flapping VIP, aggressive health check, missing fencing)

Course content

12 lessons · 11 h 5 min
Section 1 — Fundamentals1 lesson · 40 min
  1. Putting numbers on availability: SLA, MTTR, RTO, RPO and single points of failure Preview40 min
Section 2 — Load balancing4 lessons · 3 h 10 min
  1. HAProxy: frontends, backends, L4 and L7 load balancing 45 min
  2. HAProxy: health checks, session affinity and algorithms 50 min
  3. HAProxy: TLS termination, statistics page and admin socket 50 min
  4. Nginx as a load balancer: upstream, passive failure detection and TCP streams 45 min
Section 3 — Failover and clusters3 lessons · 2 h 35 min
  1. Keepalived and VRRP: a virtual IP that follows the service 50 min
  2. Pacemaker and Corosync: cluster architecture and quorum 50 min
  3. Fencing (STONITH), resources and constraints in Pacemaker 55 min
Section 4 — Data and operations3 lessons · 2 h 40 min
  1. Replicating data: DRBD, PostgreSQL and MariaDB 55 min
  2. Failovers, recovery tests and maintenance without downtime 50 min
  3. Real incidents: split-brain, flapping VIP, overly aggressive health check, missing fencing 55 min
Section 5 — Project1 lesson · 2 h
  1. Final project: making the MasterLinux shop highly available 2 h

Requirements

  • The “Fundamentals” course (included) or solid hands-on experience with the Linux command line
  • Able to administer a Debian or Rocky server (systemd services, IP networking, logs)
  • Basic knowledge of web servers (Nginx or Apache) and databases

Who is this course for?

  • System and network administrators running critical services
  • Operations and on-call engineers at IT service companies or hosting providers
  • Students preparing for a first job as an infrastructure or DevOps engineer