Key Concepts
Cluster
A Kubernetes cluster with control plane and worker nodes.
Build
Automated provisioning that installs Kubernetes components on nodes.
Reset
Complete cluster teardown that removes all configurations.
Upgrade
Rolling update of Kubernetes and component versions.
Required Permissions
| Action | Permission |
|---|---|
| View clusters | iam:project:infrastructure:kubernetes:read |
| Create/Edit clusters | iam:project:infrastructure:kubernetes:write |
| Delete clusters | iam:project:infrastructure:kubernetes:delete |
How to Create a Cluster
Configure Basic Settings
- Cluster Name - Unique identifier
- Cluster Domain - DNS domain (e.g.,
cluster.local)
Configure Versions
- K8s Version - Select Kubernetes version
- Runtime - Container runtime (containerd)
- Runtime Version - Specific runtime version
Configure Networking
- Pod Network - CIDR for pod IPs (e.g.,
10.244.0.0/16) - Service Network - CIDR for services (e.g.,
10.96.0.0/12) - CNI Plugin - Network plugin (Calico, Cilium, etc.)
Configure Options
- Proxy Mode - iptables (default) or ipvs
- CGroup Driver - systemd (recommended) or cgroupfs
- Metrics Server - Enable for resource monitoring
Creating a cluster only saves the configuration. Run the Build operation to provision Kubernetes on the nodes.
How to Build a Cluster
Building installs Kubernetes components on all configured nodes.Review Node Readiness
The wizard checks all nodes for:
- SSH connectivity
- OS compatibility
- Disk, memory, and CPU availability
- Required ports
How to Reset a Cluster
Resetting removes all Kubernetes configurations and returns nodes to clean state.Review Impact
The wizard shows affected resources:
- Running workloads (pods, deployments)
- Storage (PVs, PVCs)
- Network resources (services, ingresses)
- Configuration (configmaps, secrets)
Complete Safety Checklist
Acknowledge that:
- Data has been backed up
- Workloads will be terminated
- Storage data will be lost
- Action is irreversible
How to Upgrade a Cluster
Upgrade is only available for clusters in Ready state.
How to Delete a Cluster
Troubleshooting
Build fails with node readiness errors
Build fails with node readiness errors
- Check SSH connectivity to all nodes
- Verify nodes meet minimum resource requirements
- Ensure required ports are not in use
- Check OS compatibility
Cluster shows Not Ready after build
Cluster shows Not Ready after build
- Check operation logs for errors
- Verify network connectivity between nodes
- Ensure CNI plugin installed correctly
- Check control plane component logs
Cannot connect to cluster
Cannot connect to cluster
- Verify API server is running
- Check network connectivity to master nodes
- Ensure kubeconfig is correct
- Verify certificates are valid
Reset operation hangs
Reset operation hangs
- Check operation logs for specific errors
- Verify SSH connectivity to all nodes
- Some resources may have finalizers preventing deletion
FAQ
What's the difference between creating and building?
What's the difference between creating and building?
Creating saves the cluster configuration. Building actually provisions Kubernetes on the nodes. You must create first, then build.
Can I change Kubernetes version after building?
Can I change Kubernetes version after building?
Use the Upgrade operation to change versions. Major version changes may require reset and rebuild.
What happens to data when I reset?
What happens to data when I reset?
All data is permanently deleted including PersistentVolumes, ConfigMaps, Secrets, and container data.
Should I use iptables or ipvs proxy mode?
Should I use iptables or ipvs proxy mode?
iptables is the default and works for most clusters. ipvs provides better performance for large clusters with 1000+ services.
What is Strict ARP mode?
What is Strict ARP mode?
Required for MetalLB Layer 2 mode. Prevents nodes from responding to ARP requests for IPs they don’t own.