karpenter: NodePool: Parse CPU quantities as cores

[RajPrakash681]

Summary

status.resources.cpu and spec.limits.cpu on a Karpenter NodePool are Kubernetes resource quantities, so they can carry a suffix. Both NodePool views read them with parseInt, which stops at the first non-digit: parseInt('1750m') is 1750, so 1.75 cores is displayed and charted as 1750 cores. parseInt also truncates a plain decimal, turning 1.5 cores into 1.

Karpenter reports these values in milli-cores whenever the amount is not a whole number of cores, which is the normal case for a NodePool that has scaled up partially. With spec.limits.cpu: 8 and status.resources.cpu: 1750m, the list column read 1750/8 and the percentage bar pinned at full.

Memory in the same two files was already correct — it goes through parseRam(). Only CPU used parseInt.

Related Issue

Fixes #1243

Changes

• Added karpenter/src/helpers/parseCpu.tsx, converting n, u and m quantities to cores and returning 0 for anything unparseable. It mirrors the shape of the existing parseRam helper next to it, and the conversion matches parseCpuQuantity in the kubeflow plugin.

  On why this is a local helper rather than the SDK's parseCpu from @kinvolk/headlamp-plugin/lib/lib/units: in the pinned headlamp-plugin 0.14.0 that function parses with parseInt, so parseCpu('1.5') / TO_ONE_CPU is 1 rather than 1.5, and an unparseable value yields NaN instead of 0, which would feed NaN straight into PercentageBar. Using it here would trade this bug for a smaller one. Happy to switch to the SDK function instead once a release carries the parseFloat version, if maintainers would rather wait for that.

• Added karpenter/src/helpers/parseCpu.test.ts covering milli-cores, micro-cores, nano-cores, unsuffixed and decimal quantities, and the unparseable case.

• NodePool/List.tsx and NodePool/Details.tsx now read CPU through parseCpu instead of parseInt.

Steps to Test

1. cd karpenter && npm install && npm run test — the four new parseCpu cases pass.
2. npm run lint && npm run tsc — both clean.
3. Against a cluster running Karpenter, open Karpenter → NodePools with a NodePool whose status.resources.cpu is reported in milli-cores, for example 1750m, and spec.limits.cpu set to 8.
4. The CPU column reads 1.75/8 and the bar sits near 22%, instead of 1750/8 at full.
5. Open that NodePool's details page: CPU reads 1.75 of 8 instead of 1750 of 8.