Self-Hosted Supabase Docker RAM: 12-Service Audit — SupaScale Docker Guide
SupaScale explains why an untouched git clone of supabase/docker reserves over 4.3 GB of RAM before serving user traffic — and how we cap container memory cleanly.
SupaScale minimum floor: 8 GB RAM • Sweet-spot tier: 16 GB RAM (Public vendor rates retrieved 2026-11-12).
Three Erlang/BEAM + Lua runtimes (Logflare + Vector ~1.52 GB, Kong Nginx/Lua ~440 MB, Realtime + Supavisor ~730 MB) reserve ~2.7 GB before PostgreSQL even allocates shared_buffers. On a 2 GB or 4 GB VPS, Linux oom-killer terminates postgres pid 1 during autovacuum.
SupaScale Deterministic Memory Model Table: Self-Hosted Supabase Docker RAM: 12-Service Audit
| Supabase Docker Subsystem | Baseline Idle Reserve | 50K MAU Projected Reserve | Memory Role |
|---|---|---|---|
| Logflare + Vector Analytics (BEAM + Rust) | 1,520 MB | 1,700 MB | OOM Culprit #1 |
| PostgreSQL 15 + pgvector (Core DB) | 540 MB | 1,620 MB | Core ACID + Cache |
| ImgProxy (Go + libvips WebP Arena) | 70 MB | 670 MB cap | Burst Spike #2 |
| Kong + Supavisor + Realtime + GoTrue + REST | 1,420 MB | 2,120 MB | Gateway & Pool |
SupaScale docker-compose.override.yml — Deterministic Memory Guards
services:
db:
mem_limit: 4096m
command:
- postgres
- -c
- shared_buffers=1536MB
- -c
- effective_cache_size=4608MB
- -c
- work_mem=16MB
analytics:
mem_limit: 1536m
environment:
ERL_FLAGS: "+MBas aobf +MBlmbcs 512"
imgproxy:
mem_limit: 768m
environment:
IMGPROXY_CONCURRENCY: "4"
MALLOC_ARENA_MAX: "2"
kong:
mem_limit: 640m
environment:
KONG_NGINX_WORKER_PROCESSES: "2"1. SupaScale Audit: Why Outdated 2GB VPS Guides Crash Modern Supabase Docker
Older community guides often claim self-hosted Supabase fits on a $4 2 GB cloud droplet. That assumption held in 2022 when Supabase shipped 6 lightweight containers. Today, upstream supabase/docker orchestrates 12 logical services (13 containers) including three Elixir/Erlang BEAM clusters (Supavisor, Realtime, and Logflare) plus Rust Vector and OpenResty Kong. In our deterministic SupaScale sizing model, we sum those baseline runtime heaps to ~4.35 GB before user load even begins.
2. SupaScale Formula for Sizing PostgreSQL shared_buffers Beside 11 Sidecars
Textbook PostgreSQL documentation suggests assigning 25% of total host RAM to shared_buffers. On a single-node Supabase Docker server, we subtract active sidecar reservations first: on an 8 GB Hostinger KVM 2 running all 12 services, cap PostgreSQL shared_buffers at 1,536 MB. Stepping up to a 16 GB KVM 4 node unlocks 4,096 MB of shared_buffers plus ~6 GB of Linux filesystem page cache.
3. SupaScale Worker Caps for Kong LuaJIT and Deno Edge Runtime
By default, Kong spawns one OpenResty Nginx worker per visible vCPU core, duplicating LuaJIT schema caches in memory. We recommend setting KONG_NGINX_WORKER_PROCESSES=2 in your Compose override to trim Kong RAM by ~35% on 4-to-8 vCPU hosts.
SupaScale Matched KVM & VDS Tiers for Self-Hosted Supabase Docker RAM: 12-Service Audit
Hostinger KVM NVMe VPS (1-Click Docker / Coolify + Supabase) — SupaScale
From $4.99/mo (~127k)
12–24 mo KVM 2 (8GB) or KVM 4 (16GB) locks lowest rate + stacks Extra 20% partner discount.
Check Hostinger KVM VPS Specs for Self-Hosted Supabase Docker RAM: 12-Service Audit on SupaScale →Vultr 3.5GHz+ High Frequency NVMe (Low-Latency pgvector & Realtime) — SupaScale
From $24/mo (~612k)
Hourly billing lets you benchmark HNSW pgvector recall on 16GB/32GB HF nodes for cents.
Compare Vultr High Frequency Options for Self-Hosted Supabase Docker RAM: 12-Service Audit on SupaScale →SupaScale rate citation: Public vendor tariffs retrieved 2026-11-12; promotional and renewal prices change — verify exact checkout pricing before ordering.