April 7, 2026
Server went brrr, disk went nope
Running Out of Disk Space in Production
Downloads Go Live, Server Chokes: Fans Roast Analytics, Praise “Ballast” Hack
TLDR: A small download server crashed under demand and filled its disk; the dev freed space and moved files to new storage to recover. Commenters rallied with survival hacks (delete “ballast” files), criticized heavy analytics, endorsed lean tools, and swapped “no space left” war stories—turning a outage into a teachable meme.
A feel‑good launch turned into a storage horror flick: a tiny 40GB server tried to serve 2.2GB downloads to a stampede of customers and promptly maxed out. Emails started bouncing with “no storage” errors, the app wheezed, and the dev went full firefighter—purging logs, trying to tidy the system’s software folder, and finally bolting on new disk space to keep the downloads flowing.
But the real show was in the comments. Team Practical showed up first: flanfly dropped the “ballast files” hack—keep a few dummy gigabytes around so you can delete them in emergencies and buy time. entropie came armed with tools, hyping gdu as a faster, friendlier way to find space hogs. Then Team Minimal went for blood: huijzer blamed the bulk on Plausible analytics and pitched leaner, log‑based tracking like GoAccess instead. Meanwhile, brunoborges delivered a classic industry war story about an Oracle setup down for days… all because of “No disk space left.”
And for dessert, the line “Note: this was written fully by me, human” got a chuckle—because even the post felt like it was gasping for air. The vibe? Storage Jenga meets launch‑day chaos, with a chorus of “been there” and “ditch the bloat.”
Key Points
- •A Hetzner NixOS server (4GB RAM, 40GB disk) hosting Kanjideck downloads quickly hit 100% disk usage after launch.
- •Errors included SMTP 452 “Insufficient system storage,” indicating disk exhaustion affecting services like email.
- •Main disk consumers identified were Plausible’s ClickHouse database (~8.5GB) and the Nix store (~15GB).
- •Initial cleanup (nix-collect-garbage) failed due to no space; freeing journal logs enabled garbage collection but ClickHouse truncation still failed.
- •With no instance upgrade available, the Nix store was moved to a new attached ext4 volume following NixOS documentation.