
You searched for "node VPS" or "node hosting" and want to know what you are actually renting: how much RAM a Node.js app needs, which Node.js version belongs on it in November 2026, and how to keep the process running after you close SSH. This page answers those with the numbers from the Node.js docs, the official release schedule and the error messages a small VPS really prints.
Quick answer: A Node VPS is an ordinary Linux VPS on which you install and run Node.js yourself, for an API, a Discord bot, a Next.js site or background workers. Size it from what your process really uses plus room for the OS, nginx and any database: the Node.js docs take a 2 GiB machine as their example and suggest a heap limit of 1536 MiB there, which leaves a quarter of the RAM for everything else. Run Node.js 26, the Active LTS line since 28 October 2026 (26.13.0 as of 4 November 2026), installed with nvm as a normal user, keep it alive with PM2 or systemd, and put nginx with HTTPS in front.
What "Node VPS" and "Node Hosting" Mean
Most people who search "node VPS" mean a virtual server for Node.js. On a VPS with root access you pick the Node.js version, the process manager, the database and the reverse proxy, and you also own the updates.
The second meaning comes from game panels. In Pterodactyl, a node is a machine you add under Nodes in the admin area. It runs Wings, which the project calls "Pterodactyl's server control plane", and Wings starts and controls the game servers. "Node hosting" in that sense means renting a VPS to add as a node; our Pterodactyl Panel vs Wings guide covers that. The rest of this page is about Node.js.
How Much VPS a Node.js App Needs
The Node.js CLI docs give one concrete number. On --max-old-space-size: "On a machine with 2 GiB of memory, consider setting this to 1536 (1.5 GiB) to leave some memory for other uses and avoid swapping." That is 75% for the JavaScript heap and 25% for the rest.
Two details matter. The flag caps V8's old memory section, not the whole process: the docs for process.memoryUsage() define rss as the memory of the process "including all C++ and JavaScript objects and code", and Node.js Buffers are reported under external, apart from the heap figures. And the limit applies per process. So measure rss of your running app, then plan with the same ratio for all Node.js processes together. Our sums:
| VPS RAM | 75% for all Node.js heaps together | Left for the OS, nginx and a database |
|---|---|---|
| 2 GiB | 1536 MiB (the docs' own example) | 512 MiB |
| 4 GiB | 3072 MiB | 1024 MiB |
| 8 GiB | 6144 MiB | 2048 MiB |
| 12 GiB | 9216 MiB | 3072 MiB |
A worked example: three PM2 cluster instances on a 4 GiB VPS share the 3072 MiB, so each gets --max-old-space-size=1024. If PostgreSQL or MySQL runs on the same VPS, take its memory off the heap budget first. Don't set the limit tight to save RAM either: the docs note that as memory use approaches the limit, V8 spends more time on garbage collection.
Set it on the command line, through NODE_OPTIONS (the docs list this flag as allowed there), or as a share of RAM with the percentage flag, which the Node.js 22, 24 and 26 docs all describe and which wins when both are set:
node --max-old-space-size=1536 server.js
NODE_OPTIONS=--max-old-space-size=1536 npm run build
node --max-old-space-size-percentage=75 server.js
Builds use the same RAM as your running apps. The Next.js memory guide suggests next build --experimental-debug-memory-usage (from 14.2.0) to print heap use during the build, turning off source maps with productionBrowserSourceMaps: false and experimental.serverSourceMaps: false, and, only when a CI runner already checks types, skipping the TypeScript step with typescript.ignoreBuildErrors: true.
Which Node.js Version to Run in Production
The Node.js releases page is direct: "Production applications should only use Active LTS or Maintenance LTS releases." The lines on 4 November 2026, from the release schedule and nodejs.org/dist:
| Version | Codename | Status | Latest release | End of life |
|---|---|---|---|---|
| 26 | Lithium | Active LTS since 28 Oct 2026 | 26.13.0 (28 Oct 2026) | 30 Apr 2029 |
| 24 | Krypton | Maintenance LTS since 20 Oct 2026 | 24.22.2 (27 Oct 2026) | 30 Apr 2028 |
| 22 | Jod | Maintenance LTS since 21 Oct 2025 | 22.23.4 (20 Oct 2026) | 30 Apr 2027 |
| 20 | Iron | End of life | 20.20.2 (24 Mar 2026) | 30 Apr 2026 |
A new server gets 26. An app on 24 is supported until April 2028, so plan the move without rushing. 22 has less than six months left. 20 gets no more fixes, so upgrade it before you move it to a new server. Check the engines field of your main dependencies first; for discord.js, see Node.js 22 vs 24 vs 26 for discord.js.
Setting Up a Node VPS, in Order
- As root on Ubuntu or Debian, create a normal user with
adduser deployand install nginx with the package manager. - Log in as that user and install Node.js with the nvm commands from the nodejs.org download page. nvm is "designed to be installed per-user", and with it
npm install -gneeds no sudo.
curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.40.8/install.sh | bash
\. "$HOME/.nvm/nvm.sh"
nvm install 26
node -v # v26.13.0 on 4 November 2026
- Keep the app running with PM2 (7.1.2 since 3 November 2026; the full setup is in our PM2 production guide) or with a systemd unit such as
/etc/systemd/system/myapp.service:
[Unit]
Description=My Node.js app
After=network.target
[Service]
User=deploy
WorkingDirectory=/home/deploy/app
Environment=NODE_ENV=production
Environment=PORT=3000
ExecStart=/home/deploy/.nvm/versions/node/v26.13.0/bin/node server.js
Restart=on-failure
[Install]
WantedBy=multi-user.target
Load it with sudo systemctl daemon-reload and sudo systemctl enable --now myapp, and read its logs with journalctl -u myapp. systemd only finds programs by name in standard folders such as /usr/bin, so the nvm binary in your home folder needs its full path. That path contains the version: after an upgrade with nvm, change ExecStart, or with PM2 run pm2 unstartup and then pm2 startup, as the PM2 docs say, so the server boots the new binary.
- Let the app listen on
127.0.0.1:3000and put nginx with HTTPS in front. The Next.js deploy guide and several apps on one VPS show the config.
Errors You Will See on a Small VPS
| What you see | What it means | What to do |
|---|---|---|
FATAL ERROR: Reached heap limit Allocation failed - JavaScript heap out of memory | That Node.js process hit its own heap limit and aborted | With free RAM, raise --max-old-space-size; without it, look for a leak or use more RAM |
Killed in the shell or signal SIGKILL from npm, with Out of memory: Killed process 12345 (node) in the kernel log | The Linux OOM killer stopped the process because the whole VPS ran out of RAM | Lower the heap limits until they fit, stop other services during builds, build elsewhere or add RAM |
Error: listen EADDRINUSE: address already in use :::3000 | Another process already listens on that port, often an old copy of your app | Find it with ss, stop it or pick another port |
Error: listen EACCES: permission denied :::80 | Ports below 1024 need root or CAP_NET_BIND_SERVICE | Listen on 3000 and let nginx serve 80 and 443 |
connect ECONNREFUSED 127.0.0.1:5432 | Nothing listens there: the database or API is down or on another port | Check that service and the host and port in your config |
502 Bad Gateway from nginx | nginx got no valid response from your app | Check the app first, then our 502 guide |
The heap error and the kernel kill look alike but need different fixes: a higher heap limit can cure the first when RAM is free, and only makes the second more likely. The name in brackets in the kernel line is the process that was killed, such as node or next-build. :::3000 means port 3000 on ::, where Node.js listens when you give no host and IPv6 is available. The commands for the checks:
sudo journalctl -k | grep -i "out of memory"
sudo ss -ltnp | grep ':3000'
pm2 status
systemctl status myapp
If PM2 itself keeps restarting the app, this PM2 restart guide goes through the causes.
Bot Hosting, a VPS or a Platform
| You run | Good fit | Prices as of 4 November 2026 |
|---|---|---|
| One Discord bot, no Docker | Discord bot hosting | Growth 512 MB €0.50, Pro 1 GB €1, XL 2 GB €2 a month |
| Several apps, your own database server, Docker or nginx | A VPS | Standard S 3 vCores, 4 GB, 60 GB €4.99; M 4/8/80 €8.99; L 6/12/120 €12.99. Premium on Ryzen 9 9950X from XS 2 vCores, 2 GB, 20 GB at €4.20 (€2.10 per GB) |
| Next.js with git push deploys and no server work | A managed platform | See Vercel vs a VPS |
The bot prices are the 6 month rate (€3, €6 and €12 per 6 months), with a €0.50 setup fee on 6 and 12 month terms. Space-Node VPS plans are KVM with full root access, dedicated RAM and shared vCores, Linux images (Ubuntu, Debian, AlmaLinux, Rocky Linux) and an EU data center. They are unmanaged: you run the OS and the software, and backups are not in the plan price but available on request through a ticket.
FAQ
What is a Node VPS?
A Linux VPS on which you install Node.js yourself and run your own apps: an API, a Discord bot, a Next.js site or workers. You get root access and choose the version, process manager and database. In game panels, a node means something else: a machine that runs Pterodactyl's Wings.
How much RAM does a Node.js app need?
Measure it: rss from process.memoryUsage() shows what your process really uses. Then plan like the Node.js docs do, a 1536 MiB heap limit on a 2 GiB machine, leaving a quarter for the OS, nginx and a database. If you build on the server, leave room for the build too.
Which Node.js version should I use in production?
Node.js 26 LTS, 26.13.0 as of 4 November 2026. Node.js 24 is in maintenance until 30 April 2028 and 22 until 30 April 2027. Node.js 20 reached end of life on 30 April 2026.
PM2 or systemd?
Both restart a crashed app and start it at boot. PM2 adds cluster mode across CPU cores, log handling and pm2 reload, which PM2's cluster mode docs call a "0-second-downtime reload". systemd is already on the server and needs no npm package. With nvm, both need a change after a Node.js upgrade.
Why does npm run build get killed on my VPS?
If the output ends in Killed or npm shows signal SIGKILL, the kernel's OOM killer stopped the build because the VPS ran out of RAM. sudo journalctl -k | grep -i "out of memory" confirms it. A higher --max-old-space-size does not help here: stop other processes during the build, build in CI and copy the result, or use more RAM.
Do I need root to host Node.js?
Only for setup: installing nginx, writing a systemd unit, opening the firewall. Node.js itself should run as a normal user. Ports below 1024 need root or CAP_NET_BIND_SERVICE, which is why the app listens on 3000 and nginx serves 80 and 443.
Sources
All opened on 4 November 2026:
- Node.js release index (nodejs.org/dist/index.json)
- Node.js release schedule (nodejs/Release schedule.json)
- Node.js Releases
- Node.js CLI docs: --max-old-space-size
- Node.js CLI docs: --max-old-space-size-percentage
- Node.js docs: process.memoryUsage()
- Node.js docs: common system errors
- Node.js docs: server.listen()
- Node.js source v26.13.0: node_errors.cc
- Node.js source v26.13.0: lib/internal/errors.js
- Download Node.js
- nvm releases and nvm README
- PM2 releases
- PM2 docs: startup script
- PM2 docs: cluster mode
- systemd.service manual
- bind(2) manual and capabilities(7) manual
- RFC 9110, 502 Bad Gateway
- Next.js docs: memory usage
- Pterodactyl Wings README