Kubernetes Resource Request/Limit Calculator

How it works

Each workload row multiplies its per-pod CPU (in millicores) and memory (in Mi) request by its replica count to get aggregate totals, then divides the node's total CPU/memory (converted to the same millicore/Mi units) by each workload's per-pod request to estimate how many pods of that one workload could fit on a single node — flagging whichever resource, CPU or memory, runs out first as the bottleneck.

FAQ

Why does the per-node fit number look optimistic?

Because it only divides raw requests into raw node capacity — it doesn't subtract kubelet/system-reserved capacity, other workloads already on the node, taints/tolerations, or pod anti-affinity rules, all of which reduce real-world scheduling density.

Is the "minimum nodes needed" number based on limits or requests?

Requests only. The Kubernetes scheduler places pods based on requests, not limits, so that's what determines how many nodes you actually need — limits only affect throttling/OOM behavior once a pod is already running.