Check quota

Skill kube-dc/kube-dc-public/skills/check-quota

Check Kube-DC Organization and Project quota before creating workloads, VMs, databases, Managed Clusters, public IPs, or storage.From its SKILL.md

Install

npx -y skills add kube-dc/kube-dc-public --skill check-quota

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

SKILL.md

4.5 KB, as published. Nobody here has run it

Check Quota

Quota is governed at the Organization level and can be narrowed for an individual Project. Do not embed plan names, prices, or limits in automation; read the active values from status or the billing console.

1. Confirm context and scope

kubectl config current-context
kubectl config view --minify -o jsonpath='{.contexts[0].context.namespace}{"\n"}'

A Project context should name the selected Project and set its backing namespace. Use kube-dc use {domain}/{organization}/{project} if they do not agree.

2. Read Organization quota

kubectl -n {organization} get organization {organization} -o jsonpath='{.status.quotaUsage}' | jq .

The status can contain:

cpu, memory, storage, and pods with used and hard values;
publicIPv4, counted across the Organization's Projects;
objectStorage;
accelerators when GPU profiles are configured;
lastUpdated, the observation timestamp.

CPU is normalized to cores and memory/storage to GiB. For CPU and memory, the reported pair follows whichever requests or limits axis is closest to its cap.

An empty objectStorage.used value means usage is not available from the in-cluster controller. It is not zero; consult the provider's object-storage usage surface before provisioning a large bucket.

3. Read Project quota

kubectl -n {organization} get project {project} -o jsonpath='{.status.quotaUsage}' | jq .

perProjectQuotaSet: true means an explicit project-quota ResourceQuota is active. Otherwise, the Project status reflects the inherited Organization limit. Organization usage is still the sum across Projects, so headroom shown on one Project is not reserved for it.

Compare all Projects when deciding where capacity is being consumed:

kubectl -n {organization} get projects -o custom-columns='PROJECT:.metadata.name,CPU:.status.quotaUsage.cpu.used,MEMORY:.status.quotaUsage.memory.used,STORAGE:.status.quotaUsage.storage.used,PODS:.status.quotaUsage.pods.used,UPDATED:.status.quotaUsage.lastUpdated'

4. Inspect Kubernetes enforcement state

Status is a normalized product view. When a create is being rejected, inspect the ResourceQuota objects the API server is enforcing in the Project's backing namespace:

kubectl -n {project-backing-namespace} get resourcequota
kubectl -n {project-backing-namespace} describe resourcequota

The expected sources are:

ResourceQuota	Meaning
`hrq.hnc.x-k8s.io`	Organization quota propagated into the Project
`project-quota`	Optional tighter Project cap

This view covers native Kubernetes resources such as CPU, memory, storage, pods, and configured accelerator resource names. Public IPv4 and object-storage accounting use separate platform/provider paths and do not appear as ordinary ResourceQuota dimensions.

5. Estimate the request

Add the workload's requested capacity, not only its apparent idle usage:

Deployment or StatefulSet: replica count multiplied by each container request.
VM: vCPU, memory, and requested disk size.
Database: every database instance or replica plus its storage.
Managed Cluster: control-plane and worker-pool resources.
Public exposure: each new public EIP that will be allocated.
GPU: every quota dimension reported for the selected live profile.

Leave operational headroom for rollouts, node maintenance, database failover, and autoscaling. A workload at exactly the hard limit can fail when Kubernetes temporarily creates a replacement Pod.

Troubleshoot quota rejection

A typical API error names the ResourceQuota and exhausted resource. Capture it, then inspect both scopes:

kubectl -n {organization} get organization {organization} -o yaml
kubectl -n {organization} get project {project} -o yaml
kubectl -n {project-backing-namespace} describe resourcequota

Resolve the exhausted dimension by deleting unused resources, reducing the request, moving work only when governance permits, or requesting a quota or plan change from the Organization administrator. Do not retry unchanged manifests in a loop.

Safety

Treat lastUpdated as an observation timestamp and use ResourceQuota for a current admission failure.
Do not treat an absent or empty usage field as zero.
Check shared Organization headroom as well as a Project cap.
Confirm public IPv4 and GPU entitlement separately before creating those resources.