Kubernetes Resource Calculator
Requests, limits and node sizes — how many pods actually fit, and on how many nodes.
The scheduler places pods by requests — limits never influence placement — and a node's allocatable capacity is what remains after the OS and kubelet take their share. This does the arithmetic people otherwise get wrong on a whiteboard: pods per node, nodes needed, the limiting resource, and the QoS class your requests/limits imply.
“Reserved” models what the OS, kubelet and eviction thresholds keep back — compare kubectl describe node’s Capacity vs Allocatable. 10% is a fair default; managed clouds publish their exact formulas.
14 pods fit per node — limited by cpu
by CPU: 14 · by memory: 28 · cap: 110
Allocatable per node after the 10% reserve: 7.2 cores · 28.8 GiB. The scheduler places by requests — limits never affect placement, only throttling and OOM.
Requests below limits (or limits unset): can use spare capacity, but is evicted before Guaranteed pods under memory pressure.