How much RAM do self-hosted apps really use? Six apps measured on one VPS

"How big a server do I need?" usually gets answered with guesses. These are measured numbers: each app started with its official image on this Ubuntu 24.04 server and read with docker stats once it was idle, a minute or so after its health check passed.

What that means for server size

• 1 GB: fine for the light apps (Vaultwarden, ntfy, Forgejo). Tight for all six once the OS, Docker and a reverse proxy take their share, and any spike (an n8n workflow, a Forgejo push) can hit the out-of-memory killer. Add swap if you try.
• 2 GB: comfortable for all six with room for spikes. This is the sensible default.
• 4 GB: room to add heavier apps like Nextcloud, or a larger database.

Caveats

• Idle is the floor. n8n grows with running workflows, Umami with traffic, Forgejo during clones and pushes.
• Node-based apps (Uptime Kuma, n8n, Umami) are the heavy ones here; the Rust and Go apps (Vaultwarden, ntfy) use a fraction of that.
• Numbers depend on versions. These are the pinned versions above, measured in September 2026.

Two setup details that bit during testing

• n8n crashes on first start with EACCES: permission denied, open '/home/node/.n8n/config' when its data folder is a bind mount. Cause and fix.
• Publish every port on 127.0.0.1 and put a reverse proxy in front. A plain -p 3000:3000 is reachable from the internet even with ufw on. Umami's own example compose file publishes port 3000 publicly.