huawei-cloud-cci-instance-management
Huawei Cloud CCI (Cloud Container Instance) full lifecycle management using hcloud CLI. Covers Namespace, Network, Deployment, StatefulSet, Pod creation/update/delete/status, EIPPool for public IP, logs and metrics. Two-step confirmation for all destructive operations (delete namespace cascades all resources under it). Use this skill when the user wants to operate CCI serverless containers via command line. Triggers: CCI, 云容器实例, serverless container, 容器实例, namespace, deployment, statefulset, pod, EIPPool, CCI负载, 无服务器容器, 创建容器实例, 删除容器实例, 容器状态, 容器日志
How do I install this agent skill?
npx skills add https://github.com/huaweicloud/huaweicloud-skills --skill huawei-cloud-cci-instance-managementIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill provides a set of tools and instructions for managing Huawei Cloud Cloud Container Instances (CCI). It includes a Python helper script to handle specific API limitations and follows security best practices such as mandatory user confirmation for destructive actions and standard credential management via environment variables.
- Socketpass
No alerts
- Snykwarn
Risk: MEDIUM · 1 issue
What does this agent skill do?
Huawei Cloud CCI Container Instance Lifecycle Management
Overview
Manage Huawei Cloud CCI (Cloud Container Instance) full lifecycle using hcloud CLI (KooCLI). CCI is a serverless container service — no cluster management needed, just create a Namespace, define a Network, then deploy workloads directly.
Architecture: hcloud CLI → CCI OpenAPI → Namespace / Network / Deployment / StatefulSet / Pod / EIPPool (plus a helper script for annotation keys that hcloud CLI cannot express)
Constraints and Rules
Security Rules
- Two-step confirmation: All destructive operations (delete Namespace/Network/Deployment/StatefulSet/Pod/EIPPool/PVC) and all billable creation operations require explicit user confirmation — preview command, resource details, and cost/risk warning first; execute only after user confirms. Creating an EIPPool auto-creates EIPs and bandwidth and therefore incurs recurring charges: state the cost implication before asking for confirmation, and never create it silently.
- Credential security: Never expose AK/SK values in conversation, commands, or output. Only use
hcloud configure listto check credential status (presence only). Prefer profile mode or environment variables over explicit AK/SK parameters.
Supported Scope
This skill manages only these CCI resources and their lifecycle: Namespace, Network, Deployment, StatefulSet, Pod, EIPPool, plus read-only status/log/event/quota queries for them.
Explicitly out of scope — decline these requests, say why, and point the user at the matching skill or the Huawei Cloud console. Do not attempt them and do not invoke another skill on the user's behalf:
| Out-of-scope request | Where it belongs |
|---|---|
| CCE clusters, nodes, node pools | huawei-cloud-cce-* skills / CCE console |
| ECS / BMS instances, disks | huawei-cloud-ecs-* skills / ECS console |
| RDS, DDS, DCS databases | corresponding database skill / console |
| ELB, SFS, OBS, DNS, CDN resources | networking/storage skills / console |
| Creating VPC, subnet or security group themselves | huawei-cloud-vpc-* skills / VPC console |
| Billing, invoices, cost breakdown | Billing Center in the console |
kubectl workflows | huawei-cloud-kubectl-cce-installer |
This skill does not manage the lifecycle of other Kubernetes objects that CCI may also expose — Service, Ingress, ConfigMap, Secret, PVC, Job, Volcano Job, RoleBinding — and kubectl is out of scope. CCI does offer some write operations for those objects, but using them is not part of this skill, with one documented exception: the confirmation-gated cleanup of PVCs left behind after deleting a StatefulSet (see references/task-statefulset-management.md). Read-only inspection of ConfigMap, Secret, Service, ResourceQuota, Event, Pod status and logs is in scope and is exactly what the diagnosis playbooks use — see references/diagnosis-playbooks.md.
Resource Constraints
- Namespace flavor annotation is mandatory: Every namespace must carry the flavor annotation (value:
general-computingorgpu-accelerated). Omitting it fails creation, and the hyphenated key form only takes effect forgeneral-computing— see hcloud CLI Constraints. - limits must equal requests: CCI enforces
resources.limits == resources.requests. Mismatch causes "limit and request doesn't equal" error. Set both to the same values (e.g.,500m/1Gi). - Network must precede workloads: Pod/Deployment/StatefulSet creation fails or stays Pending if no Network exists in the namespace. Always create Network before deploying workloads.
- VPC CIDR restriction: VPC subnet CIDR must NOT be
10.247.0.0/16— CCI reserves this range for Service networking. Using it causes IP conflicts and workload creation failures. - Deletion order: Pod → Deployment/StatefulSet → EIPPool → Network → Namespace. Deleting a Namespace cascades all resources under it.
hcloud CLI Constraints
- Network creation must use Python helper script: hcloud CLI cannot pass annotation keys containing dots (
network.alpha.kubernetes.io/default-security-group). Neither dot notation nor--cli-jsonInputworks. Usescripts/cci_network_helper.py. - Namespace flavor annotation — the hyphen workaround is NOT safe for every flavor: The hyphen form (
--metadata.annotations.namespace-kubernetes-io/flavor=<value>) is normalized by CCI back tonamespace.kubernetes.io/flavoronly forgeneral-computing. Forgpu-acceleratedthe hyphen form is silently dropped and the namespace ends up asgeneral-computing, so the GPU capability is never enabled. The dotted keynamespace.kubernetes.io/flavoris the required form, but hcloud CLI cannot pass dotted annotation keys (verified: the dotted,{...}-escaped and JSON-object forms are all rejected withUSE_ERROR). Create GPU namespaces with the helper script's direct API call instead:python scripts/cci_network_helper.py create-namespace --name=<ns> --flavor=gpu-accelerated --region=<region>then confirm the canonical value withhcloud CCI readCoreV1Namespace --name=<ns>. Hyphen replacement never works for Network. - Always verify parameters with
--help: CCI has hundreds of parameters. Runhcloud CCI <Operation> --helpbefore constructing any command. The help output is the authoritative source.
Detailed breakdowns of these rules are in Security Constraints, hcloud CLI Limitations, and Precautions.
Standard Workflow
1. Create Namespace (with flavor annotation)
2. Create Network (requires Python helper script for annotation — see hcloud CLI Limitations)
3. Create Deployment / StatefulSet / Pod (run workloads)
4. Query status, view logs
5. (Optional) Create EIPPool for Pod public IP access
6. Cleanup: delete workload → delete Network → delete Namespace
Prerequisites
1. hcloud CLI Requirements (MANDATORY)
- hcloud CLI installed (version >= 7.2.2)
- Run
hcloud versionto verify installation - First-time usage:
printf "y\n" | hcloud versionto accept privacy statement hcloudmust be resolvable onPATH. If it is not (or you want to pin a specific build), setHCLOUD_PATHto the executable —scripts/cci_network_helper.pyhonours it. You can also bypass project-ID auto-detection entirely by passing--project-idto the helper.
2. Credential Configuration
hcloud CLI supports two credential modes. See references/credential-configuration.md for full details.
Quick setup (choose one):
# Mode A — Long-term AK/SK
export HUAWEI_CLOUD_AK=<your-ak>
export HUAWEI_CLOUD_SK=<your-sk>
export HUAWEI_CLOUD_REGION=cn-north-4
# Mode B — Temporary AK/SK + SecurityToken
export HUAWEI_CLOUD_AK=<your-temp-ak>
export HUAWEI_CLOUD_SK=<your-temp-sk>
export HUAWEI_CLOUD_SECURITY_TOKEN=<your-security-token>
export HUAWEI_CLOUD_REGION=cn-north-4
- Security rules: Never expose AK/SK/SecurityToken values. Use
hcloud configure listto check presence only.
⚠️ Known limitation — Python helper script credentials are independent of hcloud CLI: The Python helper script (
scripts/cci_network_helper.py) usesHW_ACCESS_KEY/HW_SECRET_KEY(and optionallyHW_SECURITY_TOKEN) environment variables for authentication, which are separate from hcloud CLI's credential source (profile orHUAWEI_CLOUD_AK/HUAWEI_CLOUD_SK). If the credentials inHW_ACCESS_KEY/HW_SECRET_KEYlack the necessary IAM permissions for CCI Network creation, the script will fail with a 403 error. Ensure these variables contain credentials with sufficient CCI permissions (e.g.,CCI FullAccess). The hcloud CLI continues using its own credential source independently — running the helper script does not affect subsequent hcloud CLI commands.
3. Validation Check
hcloud version
hcloud configure list
Security Constraints
Dangerous Operation Confirmation Mechanism
This skill strictly enforces a two-step confirmation mechanism for all destructive and billable creation operations.
All destructive operations and all billable creation operations require explicit user confirmation before execution. The process:
Step 1: Preview — Show the command, resource details, risk warning, and any cost implication
Step 2: Confirm & Execute — Only after user explicitly confirms
Operations Requiring Confirmation
| Operation | Risk Level | Description |
|---|---|---|
| Create EIPPool | 💰 Billable | Auto-creates EIPs and bandwidth — recurring charges. Give the cost warning and get explicit confirmation before creating |
| Delete Namespace | 🔴 Critical | Cascades — deletes ALL resources under this namespace (Network, Pods, Deployments, etc.) |
| Delete Network | 🟠 High | Disconnects namespace from VPC; running pods lose network |
| Delete Deployment | 🟠 High | Terminates all replicas of the workload |
| Delete StatefulSet | 🟠 High | Terminates all replicas; PVC data may be lost |
| Delete Pod | 🟠 High | Terminates the container instance |
| Delete EIPPool | 🟡 Medium | Releases public IPs allocated to pods |
| Delete PVC (StatefulSet cleanup) | 🟠 High | Destroys the claim and its volume; the data cannot be recovered. Preview with listCoreV1NamespacedPersistentVolumeClaim / readCoreV1NamespacedPersistentVolumeClaim, then delete one PVC at a time |
Credential Security
- Never expose AK/SK/SecurityToken values in conversation, commands, or output
- Never ask user to input AK/SK/SecurityToken directly in conversation
- Only use
hcloud configure listto check credential status (presence only, not values) - Prefer profile mode or environment variables over explicit AK/SK parameters
Command Format Standard
CCI follows the standard hcloud format with Kubernetes-style nested parameters:
hcloud CCI <Operation> --param=value --cli-region=<region> --cli-output=json
CCI-Specific Parameter Rules
CCI parameters follow Kubernetes API conventions — deeply nested objects with dot notation:
- Annotations use
{*}format:--metadata.annotations.namespace-kubernetes-io/flavor=general-computing - Labels use
{*}format:--metadata.labels.app=my-app - Containers array (1-based):
--spec.template.spec.containers.1.name=main --spec.template.spec.containers.1.image=nginx - Resources use
{*}format:--spec.template.spec.containers.1.resources.limits.cpu=500m - Selector matchLabels use
{*}format:--spec.selector.matchLabels.app=my-app - Namespaced operations require
--namespace: all workload operations must specify namespace
⚠️ Critical: Before constructing any CCI command, always run
hcloud CCI <Operation> --helpto verify exact parameter names. CCI has hundreds of parameters; the help output is the authoritative source.
Parameter Format Details
See references/parameter-format.md for complete CCI parameter format rules and examples.
Scenario Routing
| User Intent | Reference Document |
|---|---|
| Create/query/delete Namespace | references/task-namespace-management.md |
| Create/query/delete Network | references/task-network-management.md |
| Create/query/update/delete/scale Deployment | references/task-deployment-management.md |
| Create/query/update/delete StatefulSet | references/task-statefulset-management.md |
| Create/query/delete Pod | references/task-pod-management.md |
| Create/query/delete EIPPool | references/task-eippool-management.md |
| Query status, view logs, events | references/task-logs-and-status.md |
| Full workflow (create→run→cleanup) | references/common-workflows.md |
| All CCI operations quick reference | references/cci-operation-catalog.md |
| Troubleshooting | references/troubleshooting.md |
| Diagnosis playbooks (10 ticket categories) | references/diagnosis-playbooks.md |
| Flavors, quotas, output, debugging | references/operational-notes.md |
| IAM permissions | references/iam-policies.md |
| Verification steps | references/verification-method.md |
| Correct/error pattern comparison | references/acceptance-criteria.md |
Core Commands
Namespace
# Create namespace (general-computing flavor)
hcloud CCI createCoreV1Namespace \
--metadata.name=<ns-name> \
--metadata.annotations.namespace-kubernetes-io/flavor=general-computing \
--cli-region=<region> \
--cli-output=json
# Create namespace (gpu-accelerated flavor)
# The hyphen form above does NOT work for gpu-accelerated — hcloud CLI cannot pass the
# dotted annotation key, so use the helper script's direct API call:
python scripts/cci_network_helper.py create-namespace \
--name=<ns-name> \
--flavor=gpu-accelerated \
--region=<region>
# List namespaces
hcloud CCI listCoreV1Namespace --cli-region=<region> --cli-output=json
# Read namespace details
hcloud CCI readCoreV1Namespace --name=<ns-name> --cli-region=<region> --cli-output=json
# Delete namespace (TWO-STEP CONFIRMATION REQUIRED)
hcloud CCI deleteCoreV1Namespace --name=<ns-name> --cli-region=<region>
Network
⚠️ hcloud CLI limitation: Network creation requires a Python helper script because hcloud CLI cannot pass the annotation key
network.alpha.kubernetes.io/default-security-group(contains dots that hcloud treats as nested levels). The--cli-jsonInputapproach also doesn't work due to an hcloud bug where annotations show in--dryrunbut aren't transmitted in actual requests. See hcloud CLI Limitations below.⚠️ Credential requirement: The Python helper script uses
HW_ACCESS_KEY/HW_SECRET_KEY(and optionallyHW_SECURITY_TOKEN) environment variables for authentication. This is independent from hcloud CLI's credential source (which reads fromHUAWEI_CLOUD_AK/HUAWEI_CLOUD_SKor its profile). If the credentials inHW_ACCESS_KEY/HW_SECRET_KEYlack CCI Network creation permissions, the script will fail with 403. Ensure they have sufficient IAM permissions (e.g.,CCI FullAccess). After the script runs, hcloud CLI commands continue using their own credential source unaffected.
# Step 1: Get VPC subnet details (including neutron_network_id)
hcloud VPC ShowSubnet --subnet_id=<subnet-id> --cli-region=<region> --cli-output=json
# Step 2: Create network via Python helper script
python scripts/cci_network_helper.py create \
--namespace=<ns-name> \
--name=<network-name> \
--vpc-id=<vpc-id> \
--subnet-id=<subnet-id> \
--network-id=<neutron-network-id> \
--security-group-id=<sg-id> \
--region=<region>
# Step 3: Check network status until Active
hcloud CCI readNetworkingCciIoV1beta1NamespacedNetworkStatus \
--name=<network-name> \
--namespace=<ns-name> \
--cli-region=<region> \
--cli-output=json
# List networks
hcloud CCI listNetworkingCciIoV1beta1NamespacedNetwork \
--namespace=<ns-name> \
--cli-region=<region> \
--cli-output=json
Required Network spec fields: Network creation requires attachedVPC, subnetID, networkType, AND networkID (neutron network ID). The networkID field is REQUIRED — it is the neutron network ID obtained from hcloud VPC ShowSubnet.
Required Network annotation: network.alpha.kubernetes.io/default-security-group (the correct annotation key for CCI Network security group, NOT security-group-id). This annotation must be set to the security group ID.
Deployment
# Create deployment
hcloud CCI createAppsV1NamespacedDeployment \
--namespace=<ns-name> \
--metadata.name=<deploy-name> \
--spec.replicas=1 \
--spec.selector.matchLabels.app=<deploy-name> \
--spec.template.metadata.labels.app=<deploy-name> \
--spec.template.spec.containers.1.name=<container-name> \
--spec.template.spec.containers.1.image=<image> \
--spec.template.spec.containers.1.resources.limits.cpu=500m \
--spec.template.spec.containers.1.resources.limits.memory=1Gi \
--spec.template.spec.containers.1.resources.requests.cpu=500m \
--spec.template.spec.containers.1.resources.requests.memory=1Gi \
--cli-region=<region> \
--cli-output=json
# Scale deployment
# --spec.replicas is NOT a valid parameter; the PATCH body must be a JSON Patch array.
hcloud CCI patchAppsV1NamespacedDeploymentScale \
--name=<deploy-name> \
--namespace=<ns-name> \
--io.k8s.apimachinery.pkg.apis.meta.v1.Patch='[{"op":"replace","path":"/spec/replicas","value":<new-replicas>}]' \
--cli-region=<region> \
--cli-output=json
# Update container image (same JSON Patch form; note the 0-based array index in "path")
hcloud CCI patchAppsV1NamespacedDeployment \
--name=<deploy-name> \
--namespace=<ns-name> \
--io.k8s.apimachinery.pkg.apis.meta.v1.Patch='[{"op":"replace","path":"/spec/template/spec/containers/0/image","value":"<new-image>"}]' \
--cli-region=<region> \
--cli-output=json
# Read deployment status
hcloud CCI readAppsV1NamespacedDeploymentStatus \
--name=<deploy-name> \
--namespace=<ns-name> \
--cli-region=<region> \
--cli-output=json
StatefulSet
# Create statefulset
hcloud CCI createAppsV1NamespacedStatefulSet \
--namespace=<ns-name> \
--metadata.name=<sts-name> \
--spec.replicas=1 \
--spec.selector.matchLabels.app=<sts-name> \
--spec.template.metadata.labels.app=<sts-name> \
--spec.template.spec.containers.1.name=<container-name> \
--spec.template.spec.containers.1.image=<image> \
--spec.template.spec.containers.1.resources.limits.cpu=500m \
--spec.template.spec.containers.1.resources.limits.memory=1Gi \
--spec.template.spec.containers.1.resources.requests.cpu=500m \
--spec.template.spec.containers.1.resources.requests.memory=1Gi \
--cli-region=<region> \
--cli-output=json
# Read statefulset status
hcloud CCI readAppsV1NamespacedStatefulSetStatus \
--name=<sts-name> \
--namespace=<ns-name> \
--cli-region=<region> \
--cli-output=json
Pod
# Create pod (single container instance)
hcloud CCI createCoreV1NamespacedPod \
--namespace=<ns-name> \
--metadata.name=<pod-name> \
--spec.containers.1.name=<container-name> \
--spec.containers.1.image=<image> \
--spec.containers.1.resources.limits.cpu=500m \
--spec.containers.1.resources.limits.memory=1Gi \
--spec.containers.1.resources.requests.cpu=500m \
--spec.containers.1.resources.requests.memory=1Gi \
--cli-region=<region> \
--cli-output=json
# Read pod status
hcloud CCI readCoreV1NamespacedPodStatus \
--name=<pod-name> \
--namespace=<ns-name> \
--cli-region=<region> \
--cli-output=json
# Read pod logs
hcloud CCI readCoreV1NamespacedPodLog \
--name=<pod-name> \
--namespace=<ns-name> \
--container=<container-name> \
--cli-region=<region>
EIPPool
# Create EIPPool (for pod public IP access — auto-create EIPs)
hcloud CCI createCrdYangtseCniV1NamespacedEIPPool \
--namespace=<ns-name> \
--apiVersion=crd.yangtse.cni/v1 \
--kind=EIPPool \
--metadata.name=<eippool-name> \
--spec.amount=1 \
--spec.eipAttributes.networkType=5_bgp \
--spec.eipAttributes.ipVersion=4 \
--spec.eipAttributes.bandwidth.shareType=PER \
--spec.eipAttributes.bandwidth.size=5 \
--spec.eipAttributes.bandwidth.chargeMode=bandwidth \
--spec.eipAttributes.bandwidth.name=<bw-name> \
--cli-region=<region> \
--cli-output=json
# Read EIPPool status
hcloud CCI readCrdYangtseCniV1NamespacedEIPPoolStatus \
--name=<eippool-name> \
--namespace=<ns-name> \
--cli-region=<region> \
--cli-output=json
EIPPool required fields: --apiVersion=crd.yangtse.cni/v1 and --kind=EIPPool are mandatory. spec.eipAttributes.networkType is required. The only real values are 5_telcom (China Telecom), 5_union (China Unicom), 5_bgp (dynamic BGP), 5_sbgp (static BGP) and 5_ipv6 — any other value, including the dedicated-load-balancing value shown in older revisions of this document, does not exist in any region and is rejected by the admission webhook. Which values a region accepts differs (for example cn-north-4 supports only 5_bgp and 5_sbgp); always confirm the list for your region with hcloud EIP CreatePublicip --help, whose output enumerates the supported types per region. spec.eipAttributes.bandwidth.chargeMode and name are required when auto-creating EIPs.
💰 Billable operation: creating an EIPPool auto-creates EIPs and bandwidth, which are charged. Warn the user about the cost and obtain explicit confirmation before creating an EIPPool.
Pod EIP binding: To assign an EIPPool to a Pod, add the annotation yangtse.io/eippool=<eippool-name>. hcloud CLI cannot pass dotted annotation keys, so the hyphenated form --metadata.annotations.yangtse-io/eippool=<eippool-name> is the only CLI-expressible option; CCI stores hyphenated keys verbatim, so verify the binding after creation with hcloud CCI readCoreV1NamespacedPod --name=<pod> --namespace=<ns>, and fall back to a direct API call or the console if the annotation was not recognised.
VPC/Subnet Prerequisites
A Network maps to an existing VPC subnet, so before creating one you must supply the VPC ID, the subnet ID and the neutron network ID (networkID is mandatory):
hcloud VPC ListVpcs --cli-region=<region> --cli-output=json
hcloud VPC ListSubnets --vpc_id=<vpc-id> --cli-region=<region> --cli-output=json
hcloud VPC ShowSubnet --subnet_id=<subnet-id> --cli-region=<region> --cli-output=json # neutron_network_id comes from here
⚠️ The VPC/subnet CIDR must NOT be
10.247.0.0/16— CCI reserves that range for Service networking. Full details: references/task-network-management.md.
Namespace Flavors, Quotas, Output and Debugging
Flavor values (general-computing / gpu-accelerated), per-namespace resource quotas, JSON/table
output, JMESPath filtering, the global hcloud --debug switch and best practices live in
references/operational-notes.md.
Parameter Confirmation
Before executing any CCI operation, confirm these parameters:
| Parameter | Required | Description | Source |
|---|---|---|---|
--namespace | Yes | CCI namespace name | Existing or newly created |
--cli-region | Yes | Huawei Cloud region ID | HUAWEI_CLOUD_REGION or config |
--metadata.name | Yes | Resource name | User specified |
| Flavor annotation | Yes (Namespace) | general-computing or gpu-accelerated | User choice |
| VPC/Subnet ID | Yes (Network) | From VPC ListVpcs / VPC ShowSubnet | Query existing resources |
| neutron_network_id | Yes (Network) | From VPC ShowSubnet response | Query result |
Run
hcloud CCI <Operation> --helpbefore any CCI command to verify parameter names, then cross-reference the table above.
Precautions
See references/troubleshooting.md for detailed troubleshooting.
Quick reference:
| Issue | Cause | Quick Fix |
|---|---|---|
| Namespace creation fails | Missing flavor annotation | Add --metadata.annotations.namespace-kubernetes-io/flavor=general-computing |
| Network creation fails (400/403) | Missing VPC/subnet/annotation/networkID, or credential scope insufficient | Verify subnet/neutron IDs, security group; use Python helper; use long-term AK/SK (Mode A) |
| Pod stays Pending | No Network in namespace | Create Network first |
| 403 permission error | Insufficient IAM | Check references/iam-policies.md |
| Deep nested param errors | Wrong dot notation | Use --help to verify exact parameter path |
| Annotation with dots not passed | hcloud CLI limitation | Use Python helper script for Network creation |
| EIPPool creation fails (400/422) | Missing apiVersion/kind/networkType | Add all required fields (see EIPPool section) |
| limit/request mismatch | CCI requires limits == requests | Set requests same as limits (e.g., both 500m/1Gi) |
Verification Method
See references/verification-method.md for complete verification steps.
Quick checklist:
| Step | Command | Expected Result |
|---|---|---|
| Namespace | hcloud CCI readCoreV1Namespace --name=<ns> --cli-region=<region> | status.phase=Active |
| Network | hcloud CCI readNetworkingCciIoV1beta1NamespacedNetworkStatus --name=<net> --namespace=<ns> | status.phase=Active |
| Deployment | hcloud CCI readAppsV1NamespacedDeploymentStatus --name=<deploy> --namespace=<ns> | readyReplicas >= 1 |
| Pod | hcloud CCI readCoreV1NamespacedPodStatus --name=<pod> --namespace=<ns> | status.phase=Running |
Implementation Notes
- Namespace isolation: use different namespaces for different teams/projects.
- EIPPool on-demand: EIPs and bandwidth are billable — create an EIPPool only when Pod public IP access is needed, and confirm the cost with the user first.
- Best practices, flavors, quotas, output and debugging: references/operational-notes.md
hcloud CLI Limitations
⚠️ Critical: hcloud CLI has known limitations that affect CCI operations. Understanding these is essential for successful Network creation.
| Limitation | Impact | Workaround |
|---|---|---|
| Cannot pass annotation keys containing dots (.) via CLI parameters | hcloud treats dots in parameter names as nested object levels, so --metadata.annotations.network.alpha.kubernetes.io/default-security-group creates a deeply nested structure instead of a single annotation key | Use Python helper script (scripts/cci_network_helper.py) for Network creation |
--cli-jsonInput is unreliable for annotations | Annotations appear in --dryrun output but are not transmitted in real API requests; a UTF-8 BOM also breaks parsing | Use the Python helper script; if JSON input is unavoidable, save it as plain ASCII (no BOM) |
| Namespace flavor: hyphen replacement is only partial | The hyphen form is normalized back by CCI only for general-computing; for gpu-accelerated it is silently dropped and the namespace falls back to general-computing | Use scripts/cci_network_helper.py create-namespace (dotted annotation). The workaround never applies to Network |
Why the Python helper is needed: annotation keys such as network.alpha.kubernetes.io/default-security-group and namespace.kubernetes.io/flavor contain dots, which hcloud CLI interprets as nested object levels — so they cannot be passed as CLI parameters. scripts/cci_network_helper.py builds the correct API request body directly.
References
| Document | Description |
|---|---|
| task-namespace-management.md | Namespace lifecycle operations |
| task-network-management.md | Network lifecycle operations |
| task-deployment-management.md | Deployment lifecycle operations |
| task-statefulset-management.md | StatefulSet lifecycle operations |
| task-pod-management.md | Pod lifecycle operations |
| task-eippool-management.md | EIPPool operations |
| task-logs-and-status.md | Status queries and log viewing |
| cci-operation-catalog.md | Full CCI operation quick reference |
| parameter-format.md | CCI parameter format rules and examples |
| common-workflows.md | Complete workflow sequences |
| credential-configuration.md | Credential setup (long-term AK/SK & temporary AK/SK+SecurityToken) |
| iam-policies.md | IAM permission policies |
| troubleshooting.md | Error troubleshooting |
| diagnosis-playbooks.md | Read-only diagnosis playbooks for the 10 high-frequency ticket categories |
| operational-notes.md | Namespace flavors, resource quotas, output format, debugging, best practices |
| verification-method.md | Verification steps |
| acceptance-criteria.md | Correct/error pattern comparison |
| scripts/cci_network_helper.py | Python helper script for Network creation (bypasses hcloud CLI annotation limitations) |
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/huaweicloud/huaweicloud-skills/huawei-cloud-cci-instance-management">View huawei-cloud-cci-instance-management on skillZs</a>