Running Ghost on a 1GB DigitalOcean droplet resulted in crashes with OOM errors. Here's how I fixed it with swap and tuning for a stable, low-cost blog.

I've recently run out of contract work. I've been using my tech skills on my partner's small business, and it has been a refreshing change of pace. When I worked as a cloud consultant, I used my time to design secure solutions within government clouds. I followed established processes, reviewed requirements, and typically ran my ideas by a supervisor. Progress was often slow because some tasks had to be converted into work tickets and then passed to another department. Now that I'm the sole tech person for a small business, I've noticed that my attention has shifted to very different kinds of challenges: good enough business continuity and disaster recovery, cheap server performance, and custom integrations. ## Separating, Migrating and Backing Up My Ghost Websites One of my first projects was to migrate two self-hosted Ghost blogs into two smaller servers. The migration made me confident that I had a reliable backup strategy and that I could spin up either site again quickly if anything went wrong. For the first time, I was the one making the call on what "good enough" resilience looked like for my own business. On top of that, separating the two blogs meant that I could scale up resources for a specific site. Business continuity, disaster recovery, and independent scaling were topics I didn't have to pay much attention to as a consultant; it wasn't on my list of responsibilities. ## Tuning a Cheap Server to Keep Costs Low Working as a developer for a small business means working with a modest, budget-friendly server. The pre-built virtual private server (VPS) image on DigitalOcean for a Ghost CMS deploys a resource that isn't as lean as a handspun installation, so it would cost more to run. To keep costs low, it meant learning how to squeeze more performance out of a smaller server. Beyond following the official docs for installation, I set up swap space, adjusted Linux cache pressure, and tweaked swappiness. If these custom tweaks weren't in place, the server would occasionally run out of memory, which would then lead to killing the MySQL process and cause the site to be unresponsive. These are the kinds of details I never had to think of. In the end, I've enabled myself to self-host my site for half the cost. ## Adding a CDN (and Getting Fancy With Webhooks) To improve performance and reliability, I used the free tier of Cloudflare's content delivery network (CDN) in front of the website. Adding a CDN wasn't a set-it-and-forget-it setup. I had to configure cache rules with admin or member pages in mind, and for static assets to maximize cache hits. With the cache rules in place, I had to integrate custom Cloudflare Workers so that I could custom-purge cached content automatically when triggered by Ghost webhooks. All these changes ensured that content was served to visitors quickly and that it was never stale. It felt good to design something lean and efficient that made use of custom integrations. ## Freedom to Build The most significant difference from consulting? I don't need permission to try something. If I see something to improve, I dive in and make it happen. There's a certain freedom in being the one to decide what matters, to experiment, and to see the results of that work immediately. Using my tech skills for myself feels very rewarding. * * * These experiences may seem small compared to the work I used to do, but they've been rewarding in a different way. Running my own business as a developer means I'm responsible for both the big picture and the fine details. So far, I'm enjoying the chance to wear both hats. * * * ****Support My Writing ✍️**** If you enjoy what you've read and want to fuel my focus, consider leaving a tip or subscribing to my ****Coffee tier****. It's like buying me a hot drink so I can sit down, block out the noise, and keep the words flowing. Thank you for reading, and for being part of this ongoing experiment. Support Me