Kubernetes requests and limits calculator
Turn observed CPU and memory usage into sized requests, limits and a cluster node estimate.
Kubernetes requests and limits calculator
Paste one line per container in the form container;average CPU;peak CPU;average memory;peak memory — CPU in millicores, memory in mebibytes — and this tool sizes Kubernetes requests and limits for each one, then rolls the set up into a cluster-wide reservation. It answers the question every manifest review runs into: what should "resources" actually say, given what the container really uses.
The limit is always the peak reading plus a headroom percentage you control, so a burst above what was observed doesn't immediately hit a throttle or an OOM kill. The quality-of-service class decides the request: Guaranteed sets it equal to the limit, the only way Kubernetes grants that tier; Burstable derives it from the limit-to-request ratio, floored so it never sits below the average usage you gave it — an idling container still reserves enough for its typical load. Replicas multiplies every container's numbers before they're summed, and the node-size fields turn that total into how many nodes you'd need, sized on whichever of CPU or memory is the tighter fit.
Output comes as a tab-separated table, or a YAML resources fragment ready to paste into a container spec — a select box switches between them, live. Rows that don't split into exactly five semicolon-separated fields, or carry a non-numeric or negative value, are rejected with the line number so a typo can't produce a silently wrong estimate; a blank input just clears the result.
Everything runs in your browser. Container names and usage figures never leave your device — nothing is uploaded or logged. Copy the result, download it as a .txt file, or send it straight into another tool's input.