Guessing Memory Limits Leads to Incidents
Set a container memory limit too low and it gets OOMKilled without warning; set it too high and you waste node resources, blocking other pods from being scheduled. This tool takes your application type and expected concurrent users, then recommends the nearest standard tier (128Mi, 256Mi, 512Mi...) to start from.
Base Memory and Per-User Increment by Type
| App Type | Base Memory | Per User |
|---|---|---|
| Static Site | 128Mi | +0.2Mi |
| Lightweight API | 256Mi | +0.5Mi |
| General Web App (with DB) | 512Mi | +1Mi |
| ML/Data Processing | 1024Mi | +2Mi |
This Is a Starting Point, Not a Final Answer
The recommendation assumes a typical scenario. In production, always check actual memory usage with a monitoring tool and adjust both the limit and request values based on real data.
Frequently Asked Questions
It adds a per-user increment to the app type's base memory, applies a 20% buffer, then rounds up to the nearest standard tier.
The container can be OOMKilled without warning, or swapping can slow down response times significantly.
No. An oversized limit wastes node resources and can prevent other pods from being scheduled, so tuning with real usage data is ideal.