Fit your app into small memory limits (and why it matters)

On a small plan every service has a fixed memory budget. That is good for predictability and bad for apps that assume a whole server. Here is how to stay comfortably inside it.

1. Cap the Node heap

Node does not know it is in a small container unless you tell it. Leave room for the rest of the process:

ENV NODE_OPTIONS=--max-old-space-size=256

2. Do less at startup

Large JSON loads, eager caches and many database connections all spike memory right when the service starts. Load lazily and keep the connection pool small.

3. Trim what you ship

Heavy dependencies cost memory at runtime. Install only production dependencies in the final image (see the multi-stage Dockerfile tips).

4. Fewer replicas, not more

Each replica uses its own memory and counts against your plan. Scale out only when you need availability or throughput, as covered in prongo.yaml tips.

5. Watch for the signature

A service that runs fine, grows slowly and then disappears is leaking. One that dies seconds after starting simply needs more room than its limit, or less work at startup.

If you hit the ceiling for real, that is what the planned larger plans are for. Until then, lean services are cheaper to run anywhere.