Cluster requirements¶
shaide installs onto an existing Kubernetes cluster. Any cluster meeting the requirements below can run it, regardless of distribution or provider.
This chapter lists requirements only. For building a cluster, see Cluster setup.
For the provsioning machine that runs the installer, see Provisioning prerequisites.
Summary¶
| Area | Requirement |
|---|---|
| Kubernetes | 1.30+, identical version across all nodes |
| Access | kubeconfig with cluster-admin |
| CPU nodes | 1+, 4 vCPU / 16 GB RAM / 100 GB disk |
| GPU nodes | 1+, NVIDIA GPUs exposed as nvidia.com/gpu |
| GPU drivers | Pre-installed - shaide does not install them |
| Storage | Default StorageClass, ReadWriteOnce, 200 GB+ |
| Ingress | An external load balancer - cloud-native, or MetalLB on-prem |
| Egress | Not required for cluster nodes - the provisioning machine needs it |
Details¶
| Page | Covers |
|---|---|
| Compute | Nodes, GPUs, drivers, sizing |
| Storage | StorageClasses and capacity |
| Networking | Ingress, ports, DNS, TLS, egress |
| Verification | Commands to validate a cluster |
Access¶
The installer creates CRDs, namespaces and RBAC bindings, so it requires full administrative rights:
kubectl auth can-i '*' '*' # must return: yes
Kubernetes version¶
1.30 or later, consistent across nodes. Mixed-version clusters are not supported during installation.