FAQ — Kubernetes

What instance types are available?

Hikube offers three instance families for Kubernetes nodes:

Family	Prefix	vCPU:RAM ratio	Recommended use
Standard	s1	1:2	General workloads, web servers
Universal	u1	1:4	Business apps, databases
Memory	m1	1:8	Cache, analytics, in-memory processing

Each family is available in sizes from small to 8xlarge. For example: s1.small, u1.large, m1.2xlarge.

---

How does storageClass work in a Kubernetes cluster?

The storageClass chosen in the cluster manifest is replicated inside the tenant cluster. When your workloads create PVCs in the cluster, storage is provisioned with this storageClass on the infrastructure side.

The available storageClasses are: local, replicated, and replicated-async.

Feature	local	replicated / replicated-async
Replication	Single datacenter	Multi-datacenter (synchronous or asynchronous)
Performance	Faster (low latency)	Slightly slower
High availability	No (at storage level)	Yes

tip

The default recommendation for Kubernetes is replicated, which ensures data durability at the storage level.

note

Current limitation: only one storageClass can be passed to the tenant cluster. An improvement is in progress to allow passing all storageClasses and letting the client choose based on their needs.

---

What addons are available?

The following addons can be enabled on your cluster:

Addon	Description
certManager	Automatic TLS certificate management (Let's Encrypt)
ingressNginx	NGINX Ingress controller for HTTP/HTTPS routing
fluxcd	Continuous GitOps deployment
monitoringAgents	Monitoring agents (metrics, logs)
gpuOperator	NVIDIA GPU Operator for GPU workloads

Each addon is enabled in the cluster manifest:

cluster.yaml

spec:
  addons:
    certManager:
      enabled: true
    ingressNginx:
      enabled: true

---

How do I retrieve my kubeconfig?

The kubeconfig is stored in a Kubernetes Secret automatically generated when the cluster is created:

kubectl get tenantsecret <cluster-name>-admin-kubeconfig -o jsonpath='{.data.super-admin\.conf}' | base64 -d > kubeconfig.yaml

You can then use it:

export KUBECONFIG=kubeconfig.yaml
kubectl get nodes

---

How do I scale nodeGroups?

Scaling is controlled by the minReplicas and maxReplicas parameters of each nodeGroup. The autoscaler automatically adjusts the number of nodes between these bounds based on load.

To modify the limits, update your manifest and apply it:

cluster.yaml

spec:
  nodeGroups:
    workers:
      minReplicas: 3
      maxReplicas: 15
      instanceType: "s1.large"

kubectl apply -f cluster.yaml

---

How do I add GPU nodes to my cluster?

Add a dedicated nodeGroup with the gpus field specifying the desired GPU model:

cluster-gpu.yaml

spec:
  nodeGroups:
    gpu-workers:
      minReplicas: 1
      maxReplicas: 4
      instanceType: "u1.2xlarge"
      gpus:
        - name: "nvidia.com/AD102GL_L40S"
  addons:
    gpuOperator:
      enabled: true

warning

• Don't forget to enable the gpuOperator addon so that NVIDIA drivers are automatically installed on GPU nodes.
• Each node in the GPU nodeGroup consumes 1 physical GPU. A nodeGroup with minReplicas: 4 requires 4 available GPUs, with a direct impact on billing.