Skip to content

Guide

How much RAM and vCPU for n8n, Evolution API and Coolify

To run Coolify, n8n and Evolution API together on one Ubuntu VPS, the workable floor for light production is 4 vCPU and 8 GB of RAM (tight during Docker builds) and the balanced sweet spot is 6 vCPU, 12 GB of RAM and 200 GB of SSD. On a 4 GB VPS, the first Coolify build while WhatsApp webhooks are active usually triggers the Linux OOM killer and drops Evolution API or PostgreSQL.

On this page
  1. Quick sizing summary by workload
  2. Real-world usage per stack layer
  3. Four traps that exhaust RAM and disk
  4. How to measure your current VPS in one minute
  5. When to stay on one VPS vs. split across two
  6. Next steps

The math adds up fast because you are not running three isolated binaries: each tool brings its own database, cache, reverse proxy and short CPU/memory spikes during deployments and concurrent executions.

Quick sizing summary by workload

Use this table as a direct reference before provisioning a server or upgrading an existing one:

VPS workload Typical containers Workable floor Recommended headroom Recommended SSD
n8n only (with PostgreSQL and task runner) 4 2 vCPU · 4 GB RAM 4 vCPU · 8 GB RAM 50–100 GB
Evolution API only (1–5 instances + Postgres + Redis) 4 2 vCPU · 4 GB RAM 4 vCPU · 8 GB RAM 50–100 GB
n8n + Evolution API (plain Compose, no Coolify) 6–7 4 vCPU · 8 GB RAM 6 vCPU · 12 GB RAM 100–200 GB
Coolify + n8n + Evolution API (initial production) 10–12 4 vCPU · 8 GB RAM 6 vCPU · 12 GB RAM 200 GB
Coolify + n8n (queue mode) + 10+ Evolution instances 13+ 6 vCPU · 12 GB RAM 8 vCPU · 24 GB RAM 300 GB

Comparison chart of RAM usage per stack layer when running Coolify, n8n and Evolution API on the same VPS across 8 GB, 12 GB and 24 GB tiers

Real-world usage per stack layer

These numbers reflect a live stack with persistent data and active traffic rather than freshly booted empty containers:

Layer / service Idle RAM Loaded or spike RAM CPU profile Main growth drivers
Ubuntu + Docker + proxy (Traefik or Caddy) 400–600 MB 700 MB Low Unrotated logs and high TLS concurrency
Coolify (panel + Postgres + Redis + Realtime/Soketi) 800 MB–1.1 GB 1.5 GB + builds Medium (high during builds) Parallel deployments, npm run build / bun build and old images
n8n + task runner + PostgreSQL 600–900 MB 1.8–3.5 GB Medium to high Large JSON payloads, spreadsheets, PDFs/binaries, parallel runs
Evolution API + PostgreSQL + Redis 700 MB–1.0 GB 1.8–4.0 GB Medium Connected WhatsApp instances, media attachments and event volume
Coolify Docker build (transient) 0 MB 1.5–3.0 GB 2–4 vCPU for 1–3 min Next.js/Vite, Laravel or multi-stage builds without concurrency caps

Add the steady-state baseline for Coolify + n8n + Evolution API and you are already using 2.5 to 3.6 GB of RAM before triggering a single deploy or webhook burst.

Why 8 GB works, but requires discipline

On a 4 vCPU / 8 GB RAM VPS:

  • The idle stack (~3 GB) leaves roughly 5 GB available.
  • That comfortably handles 1 to 3 Evolution API instances, moderate n8n workflows and occasional Coolify deploys, as long as heavy builds do not overlap with memory-heavy workflows.
  • Once you also deploy custom apps or APIs inside the same Coolify instance, 8 GB loses its safety buffer quickly.

Why 12 GB is the balanced sweet spot for all three

On a 6 vCPU / 12 GB RAM VPS:

  • PostgreSQL retains healthy kernel page cache even while builds and webhooks overlap, keeping n8n responsive while Evolution API reconnects sessions.
  • The 2 extra vCPUs (6 instead of 4) prevent a Coolify docker build from stalling incoming WhatsApp webhook responses.
  • 200 GB of SSD absorbs Docker image layers, n8n execution logs and WhatsApp media without filling the root disk in a few weeks.

Four traps that exhaust RAM and disk

  1. Unbounded n8n execution history: without automatic pruning (EXECUTIONS_DATA_PRUNE=true and EXECUTIONS_DATA_MAX_AGE), n8n's database grows continuously and slows down routine queries.
  2. Uncapped Docker JSON logs: Docker's default json-file driver writes indefinitely to /var/lib/docker/containers/. Set max-size: "10m" and max-file: "3" in Compose or daemon.json.
  3. Accumulated Coolify build cache: every deploy leaves dangling image layers behind. Enable scheduled Docker cleanup in Coolify's server settings so the disk never hits 100%.
  4. No swap safety net or container limits: without a small swapfile (for instance 2 GB with vm.swappiness=10) and mem_limit caps on spiky containers, a single build surge kills PostgreSQL before alerts fire.

How to measure your current VPS in one minute

If your stack is already running and you want to check whether you need a larger tier or just better limits, run these four commands on the VPS:

bash
# 1. Real usable memory (check "available", not "free")
free -h

# 2. Live usage per container
sudo docker stats --no-stream --format "table {{.Name}}\t{{.CPUPerc}}\t{{.MemUsage}}\t{{.MemPerc}}"

# 3. Check if the Linux kernel already killed any process for RAM
sudo dmesg -T | grep -i -E 'oom|killed process' | tail -n 20

# 4. Root disk usage and Docker footprint
df -h / && sudo docker system df

Clear signs your server is undersized for its current workload:

  • available in free -h stays below 15% of total RAM outside peak hours.
  • dmesg shows any Out of memory: Killed process line.
  • Load average in uptime stays above the VPS vCPU count throughout business hours.

When to stay on one VPS vs. split across two

Situation Better option
Up to 5 WhatsApp instances, light/medium flows and a few deploys per week One 12 GB VPS (or 8 GB without Coolify) keeps internal networking, backups and maintenance simple
Frequent heavy builds (Next.js, Rust, Java) alongside business-critical WhatsApp traffic Split: one VPS for Coolify/builds and a second dedicated to n8n + Evolution API so builds never stall webhooks
10+ Evolution API instances or n8n in queue mode with multiple workers 24 GB VPS (8 vCPU) or move PostgreSQL / queue workers onto a dedicated server

Next steps