Kogent.Tech — Infrastructure Engineering Blog

How We Built a Self-Healing DNS Failover for $0 Using Cloudflare Workers

Wed, 06 May 2026 00:00:00 +0000

Every hosting provider has maintenance windows. Most give you notice. Some don’t. When your provider goes dark at an inconvenient time, you have two options: scramble to restore service or switch traffic to a secondary site that’s already running and already current. The first option is stressful. The second one is engineering.

This post covers how we built automated DNS failover across a fleet of production sites — our own ‘kogent.tech’ and several client sites — using Cloudflare Workers, Cloudflare KV, an on-premises server, and a Cloudflare Tunnel. The whole system costs nothing in ongoing fees. Failover takes under two minutes. Failback is automatic.

Building a Data-Sovereign Mail Server, Part 1: Classical Hardening

Wed, 29 Apr 2026 00:00:00 +0000

Running your own mail server has a reputation. The conventional wisdom is that it’s a masochistic endeavour best left to people who enjoy reading RFC errata for fun on a Sunday afternoon. Spam filters will hate you. Your IP will be blacklisted before you send your first message. You’ll wake up at 3am to a silent inbox and spend an hour discovering that a single missing SPF record cost you a week of legitimate mail.

Building a DevOps blog: Hugo, GitLab CI, and Cloudflare Pages

Tue, 31 Mar 2026 00:00:00 +0000

A DevOps blog should practice what it preaches. So when I decided to start writing about infrastructure work, I wasn’t going to click through a WordPress installer or pick a Squarespace template. The blog itself needed to be a project — version-controlled, automated, infrastructure-as-code, and deployed through a proper CI/CD pipeline.

This is the write-up of how I built blog.kogent.tech from scratch — the stack decisions, the theme build, the pipeline, and everything that broke along the way.

Migrating six WordPress sites to Kubernetes: what actually broke

Tue, 24 Mar 2026 00:00:00 +0000

Running six WordPress sites on a bare-metal Debian 11 LAMP stack is fine — until it isn’t. No secret management, no automated backups, and deployments meant SSH-ing into the server and copying files. By early 2026 the technical debt was obvious enough to justify a full migration.

The goal: migrate everything to Kubernetes, implement proper secret management with HashiCorp Vault, automate backups, and decommission the old host entirely. This post covers what that actually looked like — not the happy path, but the parts that broke and why.