Docker’s default 64 MB /dev/shm partition and unreaped renderer subprocesses crash unhardened Playwright deployments within hours. Our BotFoundry model allocates 0.78 GB per active BrowserContext group, enforces shm_size: "2gb" plus init: true, and pairs worker pools with high-RAM DDR5 KVM or Ryzen servers.
Public rates retrieved 2026-11-12 from official vendor pricing pages cited below; promotional & renewal prices may change — always verify exact specs, term length, and renewal rates on the vendor’s official checkout page.
| Concurrency Scale | vCPU | DDR5 RAM | OS / Runtime | /dev/shm & Watchdog Guard | Public Tier Match (2026-11-12) |
|---|---|---|---|---|---|
| 1–4 Contexts | 2 vCPU AMD EPYC | 4 GB DDR5 | Ubuntu 24.04 Docker | shm_size: 1gb + init: true |
Hostinger KVM 1 ($4.99/mo) |
| 8–12 Contexts | 4 vCPU KVM | 16 GB DDR5 | Ubuntu 24.04 Docker | shm_size: 2gb + Recycle 250 |
Hostinger KVM 4 ($10.49/mo) |
| 20–35 Contexts | 8 vCPU KVM / VDS | 32 GB DDR5 | Ubuntu 24.04 + BullMQ | shm_size: 4gb + MaxOld=4096 |
Hostinger KVM 8 ($19.99/mo) / InterServer 32GB |
| 60+ Stealth Contexts | 12–16 Core Ryzen | 64 GB DDR5 | Bare-Metal Ubuntu 24.04 | shm_size: 8gb + Swappiness=10 |
InterServer Ryzen 64GB ($89/mo) |
services:
pw-worker:
image: mcr.microsoft.com/playwright:v1.49.0-noble
init: true # Tini PID 1 reaps orphaned Chromium renderer zombies
shm_size: '2gb' # Eliminates 64MB /dev/shm Target Closed crashes
restart: unless-stopped
environment:
- NODE_OPTIONS=--max-old-space-size=4096
- MAX_CONCURRENT_CONTEXTS=10
deploy:
resources:
limits:
memory: 12G
command: ["node", "dist/worker.js"]Chromium’s Skia compositor and V8 IPC channels allocate shared memory frames inside /dev/shm; once a complex React storefront exceeds 64 MB, tabs throw "Target page, context or browser has been closed." Pairing shm_size: "2gb" with init: true reaps exited renderer PIDs immediately so RSS memory stays flat across weeks.
Calling chromium.launch() on every queued URL burns 280–450 MB of cold startup RAM. Launch one parent Chromium process per worker container, spawn isolated browser.newContext({ proxy }) tabs (~80–120 MB each), and recycle the parent binary cleanly every 250 completed tasks.
Intercept non-essential assets with page.route("**/*.{png,jpg,webp,mp4,woff2}", r => r.abort()) whenever visual screenshots are not required — halving CPU decode load on a $10.49/mo 16 GB Hostinger KVM 4 slice.
Serverless 15-minute execution ceilings, cold Chromium binary decompression, and $0.09/GB cloud egress tariffs make continuous scraping 5x–8x costlier than a flat-rate 16 GB–32 GB KVM VPS on Hostinger or InterServer.
No — pass --disable-gpu for HTML/JSON extraction. Only specialized WebGL canvas workflows require software SwiftShader threads, where high-core AMD EPYC or Ryzen CPUs excel.
We provision a 4 GB NVMe swapfile with sysctl vm.swappiness=10 as an emergency shock absorber while enforcing strict concurrency caps in BullMQ.
BotFoundry 4-layer RAM equation, /portable chart trimming, and Windows AutoLogon watchdog rules for 1 to 50+ MetaTrader 4/5 Expert Advisor terminals.
BotFoundry sizing formulas for Ubuntu 24.04 Freqtrade Docker bots, cgroup MemoryMax bounds, chronyd NTP sync, and isolated FreqAI LightGBM retraining.
BotFoundry 4-layer RAM guide for scaling self-hosted n8n Queue Mode with dedicated webhook receivers, Redis AOF queues, and PostgreSQL 16 pruning.
BotFoundry sizing formulas for Hummingbot V2 Controllers, single-thread 3.5GHz+ Cython loops, EMQX MQTT brokers, and TypeScript DEX Gateway containers.