Description

The render.yaml Blueprint specifies plan: starter, but the OpenClaw gateway requires more than 512MB RAM to start. The Node.js process consistently OOMs during gateway initialization.

Evidence

Tested on Render Starter plan (512MB):

1. V8 heap OOM at default (250MB): FATAL ERROR: Reached heap limit Allocation failed
2. V8 heap OOM at 384MB: Same heap OOM, peak usage ~375MB
3. Container OOM at 450MB heap: V8 survived but Render killed the container: ==> Out of memory (used over 512Mi)

The gateway needs ~450MB for V8 heap + ~80-100MB for V8 internals/Node.js runtime/OS = ~530-550MB total, exceeding the 512MB Starter plan.

Fix

Update render.yaml to use plan: standard (2GB RAM):

services:
  - type: web
    name: openclaw
    runtime: docker
-   plan: starter
+   plan: standard

Additional issues found in render.yaml

1. Missing gateway subcommand: The Dockerfile CMD ["node", "dist/index.js"] runs without the gateway subcommand, causing the CLI to print help and exit. Needs a dockerCommand override or Dockerfile fix (see Dockerfile CMD missing gateway subcommand - container exits immediately on Render #5685).

2. Port mismatch: The PORT=8080 env var is set but OpenClaw reads OPENCLAW_GATEWAY_PORT, not PORT. The gateway defaults to 18789 while Render Docker services route to 10000.

Working configuration

For reference, this Docker command works on the Standard plan:

node dist/index.js gateway --port 10000 --bind lan --allow-unconfigured

Environment

• Render Standard plan (2GB RAM) - working
• Render Starter plan (512MB) - OOM on every attempt
• OpenClaw v2026.1.30 (tag commit 76b5208)