Two servers and half of a third: building HA infrastructure for Totum

This article details the experience of building a High Availability (HA) infrastructure for the Totum low-code platform. The primary goal was to eliminate a single point of failure without significant hardware investment. The authors aimed to ensure automatic PostgreSQL failover, user file availability, and the elimination of manual DNS changes during outages. The chosen solution is an architecture dubbed 'HA for the poor,' utilizing two main servers and a lightweight witness node to maintain quorum. The entire deployment and maintenance process is automated via Ansible. The article covers technical nuances of cluster configuration and component selection, enabling business systems to remain operational even if one virtual machine fails. It serves as a practical guide for those seeking efficient ways to ensure high availability under resource constraints.