Loading
NodeGroup operation Skill for the edgewize nodegroup project. Use this whenever the user wants to query, create, update, delete, bind, unbind, or troubleshoot NodeGroup resources, including node bindi
The skill file stays in the publisher's repository.
Paste the copied command in a terminal. It installs this SKILL.md into that agent.
npx skills add kubesphere/kubesphere --skill nodegroup -a cursor -y
NodeGroup operation Skill for the edgewize nodegroup project. Use this whenever the user wants to query, create, update, delete, bind, unbind, or troubleshoot NodeGroup resources, including node binding, namespace binding, workspace binding, and deployment/config inspection for nodegroup.
NodeGroup operation Skill for the edgewize nodegroup project. Use this whenever the user wants to query, create, update, delete, bind, unbind, or troubleshoot NodeGroup resources, including node binding, namespace binding, workspace binding, and deployment/config inspection for nodegroup.
Use this skill to perform real nodegroup-related operations in the edgewize-io/nodegroup environment.
This skill should help with:
NodeGroupNodeGroupconfig/nodegroupPrefer using the bundled script scripts/nodegroup_api.py for authenticated KAPI operations. Use kubectl only as a verification or fallback tool when the API path is unavailable.
Use the bundled script:
scripts/nodegroup_api.pyAuthentication model:
/oauth/token~/.kubesphere_tokenKUBESPHERE_HOST, KUBESPHERE_USERNAME, KUBESPHERE_PASSWORD, and KUBESPHERE_TOKENSetup:
cd scripts
pip install requests
export KUBESPHERE_HOST="http://<kubesphere-host>"
python nodegroup_api.py login --username admin --password <password>
Token helpers:
python nodegroup_api.py clear-cache
python nodegroup_api.py request GET /kapis/infra.kubesphere.io/v1alpha1/nodegroups
When you need to confirm how an operation works, read these files first from the nodegroup source repository root:
pkg/kapis/infra/v1alpha1/register.gopkg/kapis/infra/v1alpha1/handler.gopkg/kapis/infra/v1alpha1/workspace_handler.gopkg/controller/nodegroup/nodegroup_controller.gopkg/constants/constants.goDo not invent unsupported operations. Base the workflow on the source repository.
Cluster-scoped resource.
Important fields:
spec.aliasspec.descriptionspec.managerstatus.statepython nodegroup_api.py ... query commands.For any write operation, follow this order:
python nodegroup_api.py nodegroup list
python nodegroup_api.py nodegroup get <name>
python nodegroup_api.py nodegroup create \
--name <nodegroup-name> \
--alias "<alias>" \
--description "<description>" \
--manager "<manager>"
python nodegroup_api.py nodegroup get <nodegroup-name>
Prefer patching for small changes.
python nodegroup_api.py nodegroup patch <name> \
--alias "<new-alias>" \
--description "<new-description>" \
--manager "<manager>"
python nodegroup_api.py nodegroup get <name>
For unsupported fields or ad hoc testing, use raw request mode.
For PATCH, send a JSON Patch array:
python nodegroup_api.py request PATCH /kapis/infra.kubesphere.io/v1alpha1/nodegroups/<name> '[{"op":"add","path":"/spec/alias","value":"<new-alias>"}]'
Only do this when explicitly requested.
python nodegroup_api.py nodegroup delete <name>
If delete hangs, inspect finalizers:
python nodegroup_api.py nodegroup get <name>
Relevant finalizer:
finalizers.nodegroups.kubesphere.ioNode binding is implemented through nodegroup APIs and reflected with label:
apps.edgewize.io/nodegroup=<nodegroup-name>Read current state:
kubectl get node <node-name> --show-labels
kubectl get nodes -l apps.edgewize.io/nodegroup=<nodegroup-name>
Operate through the bundled script:
python nodegroup_api.py bind node --nodegroup <nodegroup-name> --node <node-name>
python nodegroup_api.py unbind node --nodegroup <nodegroup-name> --node <node-name>
# Verify
kubectl get nodes -l apps.edgewize.io/nodegroup=<nodegroup-name>
kubectl get node <node-name> -o yaml
Read current state:
kubectl get ns <namespace> --show-labels
Operate through the bundled script:
python nodegroup_api.py bind namespace --nodegroup <nodegroup-name> --namespace <namespace>
python nodegroup_api.py unbind namespace --nodegroup <nodegroup-name> --namespace <namespace>
# Verify
python nodegroup_api.py nodegroup get <nodegroup-name>
kubectl get ns <namespace> -o yaml
python nodegroup_api.py bind workspace --nodegroup <nodegroup-name> --workspace <workspace>
python nodegroup_api.py unbind workspace --nodegroup <nodegroup-name> --workspace <workspace>
When an operation appears to succeed but state is wrong, check these in order:
Useful checks:
python nodegroup_api.py nodegroup get <name>
kubectl get node <node-name> -o yaml
kubectl get ns <namespace> -o yaml
kubectl get cm nodegroup-config -n kubesphere-system -o yaml
kubectl get pods -n kubesphere-system | grep nodegroup
Important constants:
apps.edgewize.io/nodegroupapps.edgewize.io/namespace-nodegroup.infra.kubesphere.io/parentinfra.kubesphere.io/ippool-If the user asks how nodegroup is deployed or configured, inspect these files from the nodegroup source repository root:
config/nodegroup/templates/nodegroup-apiserver.ymlconfig/nodegroup/templates/nodegroup-controller-manager.yamlconfig/nodegroup/templates/nodegroup-config.yamlconfig/nodegroup/values.yamlConfigMap key:
nodegroup.yamlNotable config sections:
schedulingpolicyrbacippoolIf the user asks for an operation that already has a dedicated nodegroup API, prefer the bundled script and the dedicated nodegroup API semantics over manual label hacking.
Examples:
python nodegroup_api.py bind node ...python nodegroup_api.py bind namespace ...python nodegroup_api.py bind workspace ...python nodegroup_api.py request ...Only fall back to direct manifest edits or label repair when the request is explicitly about low-level repair.
When using this skill:
nodegroup_api.py command or API path to run next