Running a Discord Bot and Tiny API on a Budget VPS
TL;DR: Always-on bots and small JSON APIs waste money on oversized clouds and die on free tiers that sleep. A modest VPS with predictable RAM keeps WebSocket connections alive without enterprise pricing. Compare affordable VPS plans sized for low but constant traffic before you overbuy CPU you will never pin.
Why bots hate serverless sleep and love boring VMs
Discord gateway connections expect a process that stays warm. Platforms that freeze containers after idle minutes force reconnect storms and missed events. A small VPS running a single Node, Python, or Go binary avoids that class of failure. You pay for continuity—the actual product bots need—more than for burst CPU.
The same box can host a companion API for dashboards, Stripe webhooks, or a simple admin UI. Keep both under one process supervisor (systemd, Docker Compose) so logs and restarts live in one place. Splitting across three free tiers usually creates three different outage modes and zero shared tracing.
Community bots also need predictable IPs for allowlists and fewer surprise hostname changes. Budget VPS instances typically keep a stable address longer than ephemeral sandbox hosts.
Sizing without cargo-culting “4 vCPU minimum”
Most community bots idle near zero CPU and need more attention to memory leaks than to core count. Start with:
- 1 shared or 1 dedicated vCPU for a single bot plus a thin API.
- 1–2 GB RAM if you use Node or Python with modest caches; 2 GB if you also run Redis or Postgres on-box.
- 20–40 GB SSD for logs, Docker images, and backups—not for media libraries.
Upgrade when RSS climbs week over week or event latency spikes during raids—not because a marketing page sold you a “pro gamer” SKU. Watch garbage collection pauses in Node and unclosed database sessions in Python; those look like “need more RAM” when they are really leaks.
Separate data from the app when you can
If the API stores user settings, prefer a managed Postgres later. On day one, SQLite or a tiny Postgres on the same VPS is fine if you schedule filesystem snapshots. Document the path to migrate off-box before the community grows past a few thousand daily actives.
Never store Discord tokens in the repository. Environment files with restricted permissions, or a minimal secrets manager, keep a leaked git history from becoming a compromised server farm.
Networking and secrets on cheap instances
Point a subdomain at the VPS, terminate TLS with Caddy or Nginx, and never expose Postgres to 0.0.0.0. Store Discord tokens and API keys carefully. Enable a firewall that allows only 80/443 and your SSH port from known IPs when practical.
Enable automatic security updates for the base OS. Budget VPS providers still expect you to patch; “cheap” is not “someone else hardens it.” Disable password SSH in favor of keys on day one—bots attract scanners like any public host.
Observability that fits a hobby budget
You do not need a full APM suite. Ship:
- Disk usage alerts before Docker fills the volume with old images.
- A cron health check that pings your API’s
/healthand notifies you on failure. - Log rotation so a chatty shard cannot erase itself with infinite journal growth.
- A weekly reboot window only if your stack actually leaks; prefer fixing leaks over ritual restarts.
Measure reconnect counts after Discord API incidents; rising reconnects with flat traffic often means memory pressure or a buggy resume handler—not a need for eight vCPUs.
Cost discipline: annual billing vs. experiment mode
Pay monthly while the bot idea is unproven. After thirty stable days, annual discounts on budget VPS listings often beat spinning another free-tier experiment that vanishes. Track the all-in cost: VPS + domain + backups. If that exceeds what the community tip jar covers and you are not building a portfolio piece, shrink scope before upsizing hardware.
Beware “unlimited bandwidth” footnotes. Bots rarely saturate links, but a compromised box serving malware will. Keep update sources trusted and fail2ban or equivalent in place.
When to leave a single VPS
Move the database off-box when backups become scary. Split the API when deploy frequency fights the bot’s uptime requirements. Add a second region only if latency or outages actually hurt members—not because architecture Twitter said so. Until then, one well-monitored affordable instance beats a half-configured Kubernetes cluster.
If you outgrow a single process, introduce a queue for heavy commands rather than instantly doubling VPS size. Queues absorb raid-night spikes more cheaply than idle cores.
Always-on community tooling thrives on boring, paid continuity. Pick a right-sized plan from Affordable VPS Server, lock tokens down, and grow resources only when metrics—not FOMO—say the box is genuinely tight.
